Author Topic: Core World approach  (Read 12171 times)

Offline Hearteater

  • Core Member
  • *****
  • Posts: 2,334
Re: Core World approach
« Reply #45 on: January 01, 2013, 03:07:16 pm »
Also, why does the Tethuda guard post have infinite range? It doesn't seem to fit. I'd rather see that on a non-brutal pick "sniper core guard post".
Be glad it does, otherwise Scout Starships wouldn't make you immune to it.  If it had Cursed Golem or even Zenith Bombard range, we would be screwed.

Offline Kahuna

  • Core Member
  • *****
  • Posts: 2,222
  • Kahuna Matata!
Re: Core World approach
« Reply #46 on: January 02, 2013, 12:11:13 pm »
How would you destroy an Ancient Shadows core guard post is the planet has a Core Eye. I just created 10 games and half of the homeworlds had a core eye. 20 homeworlds 10 had an eye. The core eyes have planetary tachyon coverage so cloaker starships don't work. Ragnarok is actually very lucky neither of the AI homeworlds have an eye.
set /A diff=10
if %diff%==max (
   set /A me=:)
) else (
   set /A me=SadPanda
)
echo Check out my AI War strategy guide and find your inner Super Cat!
echo 2592 hours of AI War and counting!
echo Kahuna matata!

Offline TechSY730

  • Core Member Mark V
  • *****
  • Posts: 4,570
Re: Core World approach
« Reply #47 on: January 02, 2013, 12:19:15 pm »
How would you destroy an Ancient Shadows core guard post is the planet has a Core Eye. I just created 10 games and half of the homeworlds had a core eye. 20 homeworlds 10 had an eye. The core eyes have planetary tachyon coverage so cloaker starships don't work. Ragnarok is actually very lucky neither of the AI homeworlds have an eye.

Yikes! I thought the issue where the multiple eye types was skewing odds of an eye was fixed.  :o

EDIT: Wait, what difficulty? Higher difficulties get more brutal picks, and although I have not done the math to be sure, if they get 4 brutal picks (which I think happens with >=9.3), 50% chance sounds about right. Lower difficulties will get less picks, drastically reducing the odds.
« Last Edit: January 02, 2013, 12:21:07 pm by TechSY730 »

Offline Kahuna

  • Core Member
  • *****
  • Posts: 2,222
  • Kahuna Matata!
Re: Core World approach
« Reply #48 on: January 02, 2013, 12:29:20 pm »
That was 10/10
set /A diff=10
if %diff%==max (
   set /A me=:)
) else (
   set /A me=SadPanda
)
echo Check out my AI War strategy guide and find your inner Super Cat!
echo 2592 hours of AI War and counting!
echo Kahuna matata!

Offline TechSY730

  • Core Member Mark V
  • *****
  • Posts: 4,570
Re: Core World approach
« Reply #49 on: January 02, 2013, 12:33:10 pm »
That was 10/10

Ah, then you were asking for it.  ;)

Still, it wouldn't hurt to do the math and compare expected values with seen values. *tries to remember stuff in my probability class I took years ago...*

Also, it would be interesting to compare these results to the results from a bunch of 7/7 and 8/8 games.

Offline chemical_art

  • Core Member Mark IV
  • *****
  • Posts: 3,952
  • Fabulous
Re: Core World approach
« Reply #50 on: January 02, 2013, 01:34:05 pm »
Kahula I tried experimenting the AI eye logic on 10/10, and in 100% of my games (of a set of 20 games) at least one AI HW had an eye. It seems to be coded that one is a given.

I'm not a fan of the new brutal picks. As been echoed, you want the last part of the game to be more intense, with you and the ai pounding each other to the death. Not you flailing around trying to take out their heavy layers of defenses. Which is why I liked the old brutal picks of CPA's and raid engines. It makes any offense a very stressful time for both sides: The AI can't stop you immediately, but the player doesn't have much time to react to a threatening attack as well.
Life is short. Have fun.

Offline Hearteater

  • Core Member
  • *****
  • Posts: 2,334
Re: Core World approach
« Reply #51 on: January 02, 2013, 01:39:36 pm »
But I prefer the newer picks because Raid Engines and CPA posts detract from the homeworld fight.  I need to back-off to deal with them, rather than focus all my attention on the true threat of the homeworld.  I mean, I've often spent 20+ hours getting to this point, I'd like to actually fight on the homeworld a little bit.

Offline chemical_art

  • Core Member Mark IV
  • *****
  • Posts: 3,952
  • Fabulous
Re: Core World approach
« Reply #52 on: January 02, 2013, 01:44:04 pm »
But I prefer the newer picks because Raid Engines and CPA posts detract from the homeworld fight.  I need to back-off to deal with them, rather than focus all my attention on the true threat of the homeworld.  I mean, I've often spent 20+ hours getting to this point, I'd like to actually fight on the homeworld a little bit.

I guess that is why there is variety.

Like for me, I it just another blob on blob fight. To paraphrase another quote, once I get to the HW, it is like I'm a child swinging, and the AI is using their long arm to hold me back, and all I can manage to do is awkward swings at the arm till I break it, then the AI is pummeled to death very quickly thereafter.
Life is short. Have fun.

Offline TechSY730

  • Core Member Mark V
  • *****
  • Posts: 4,570
Re: Core World approach
« Reply #53 on: January 02, 2013, 02:10:29 pm »
I'm not sure how much of this is due to dislike of the goal of the new picks vs. the fact that the new picks have not been through the "balance rounds" yet. I am reserving judgement about the "external effect" vs "local effect" vs "giving more local FP" styles of brutal pick posts until the balance rounds can be completed.

Also, the OPness of the core reserve respawn rate (and possibly core AI eye tachyon coverage radius and a possible spawn rate glitch with them) currently is also muddying the issue.

Kahula I tried experimenting the AI eye logic on 10/10, and in 100% of my games (of a set of 20 games) at least one AI HW had an eye. It seems to be coded that one is a given.

Although that is suspicious, sadly, a sample size of 20 games (which is 40 HWs in all) is not really a big enough sample size to give strong evidence of something strange going on.

Offline Hearteater

  • Core Member
  • *****
  • Posts: 2,334
Re: Core World approach
« Reply #54 on: January 02, 2013, 03:03:38 pm »
From the 5.036 patch notes (when brutal picks were added):
Quote
Previously it was possible for AI Homeworlds to get really brutal combinations like an AI Eye + 3 (or more) Core Raid Engine guard posts (which is pretty much the RNG saying "good game" on higher difficulties but you don't know until you scout it), or nothing really all that threatening at all. Randomization is core to AI War, but this particular bit of it could literally be the difference between a fairly easy endgame and a mathematically-impossible-to-win situation.

Now, for AI Homeworlds, it randomizes the three most brutal structures (Core Raid Engine, Core CPA, and AI Eye) separately from the rest of the structures:
  • On Diff 9+ it always gets two picks from this new set. So you could get an Eye + a Core Raid, or two CPAs, or two Eyes (which is actually kind of nice to you since they don't really stack), or whatever. But not 3 Core Raids. And not none of the above.
  • On Diff 8+ it randomly rolls (per homeworld) between one and two picks.
  • On Diff 7+ it always gets one pick.
  • Below Diff 7 it randomly rolls between zero and one picks.
From 5.046 patch notes:
Quote
When an AI gets 2 "brutal" picks for its homeworld (only happens on high difficulty), it now cannot pick 2 AI Eyes, since that's too nice.

From 6.006 patch notes:
Quote
Core Eyes, as a group, are now roughly as likely to be picked as each of the other brutal-pick AI HW structures. Each core eye is still equally likely to be picked, compared to other eyes.

So that's 6 brutal types.  On 9+ you get two on each home world.  That's 20 possible combinations, excluding double Eyes (6C2 with repetitions becomes 7C2 without, minus one for double Eyes).  Of those, only 5 contain an Eye.  So only 25% of homeworlds should have a Core Eye.  There are two homeworlds, so the chance of getting a map with no Core Eyes is 56.25%.  So with some quick math, we get  Mean of 8.75 maps with at least one Core Eye in a sample of 20 and Standard Deviation of 2.22.  So you are a touch over 5 Standard Deviations off from the expected result, or 1 in 3,488,555.

Although the sample size is small, that suggets that maybe something is off.

Offline chemical_art

  • Core Member Mark IV
  • *****
  • Posts: 3,952
  • Fabulous
Re: Core World approach
« Reply #55 on: January 02, 2013, 04:50:43 pm »
From what I remember, some time after 6.0 while not recorded there was a change. The ai got three picks, and at least from my sample one he always had an eye. This is for 10/10
« Last Edit: January 02, 2013, 05:03:45 pm by chemical_art »
Life is short. Have fun.

Offline Hearteater

  • Core Member
  • *****
  • Posts: 2,334
Re: Core World approach
« Reply #56 on: January 02, 2013, 09:59:45 pm »
Found it in 5.083:
Quote
Added roughly (there's some randomization) one "brutal pick" to each AI homeworld on difficulties 7+.
So that brings us to 56 combinations, and then we subtract out 1 for triple Eyes, and 5 for all the possible double Eye combos, leaving us with 50 total.  There are 15 combos with an Eye, so there is a 30% chance of an Eye on a given home world.  That gives us a 49% chance of no Core Eyes on either home world.  Probability gives us a Standard Deviation of 2.236 for a sample of 20.  The Mean on our sample is now 10.2.  That's still 4.38 standard deviations.  So I'd say something is fishy in Core Eyes getting seeded, or the patch notes.  Or my ability to read the patch notes.
« Last Edit: January 02, 2013, 10:03:43 pm by Hearteater »

Offline Bognor

  • Hero Member
  • *****
  • Posts: 570
Re: Core World approach
« Reply #57 on: January 04, 2013, 11:33:45 pm »
Hmmm.  Do non-core eyes ever seed on homeworlds?
Your computer can help defeat malaria!
Please visit the World Community Grid to find out how.

Offline Hearteater

  • Core Member
  • *****
  • Posts: 2,334
Re: Core World approach
« Reply #58 on: January 05, 2013, 10:59:14 am »
Not any more.  And even if they did, non-Core don't have Tachyon coverage, so that wouldn't prevent stealth ops.

 

SMF spam blocked by CleanTalk