Author Topic: (Thought Exercise) Reduce knowledge to reduce AIP scaling.  (Read 21359 times)

Offline LaughingThesaurus

  • Master Member Mark II
  • *****
  • Posts: 1,723
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #30 on: April 12, 2013, 08:03:50 am »
The rest of the game has to be tweaked in order to really comply with that, since knowledge isn't the only thing you gain from planets.
I'm kind of intrigued at the idea of maybe letting the AI get more levels of ships and more bonus ships, and having AIP have an overall lower effect (in terms of number of ships). This way, you're always fighting not just the 'occasionally more ships' but you're always getting new stuff thrown at you moreso than you may be used to. This means you would be fighting against mark III waves and the AI would have something like 5 or 6 bonus ships towards endgame... or maybe you manage a bit less than that with a low AIP game. This way, the AI's crucial evolution as the game goes on would be a bit less... completely ignored.

Offline Hearteater

  • Core Member
  • *****
  • Posts: 2,334
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #31 on: April 12, 2013, 08:48:03 am »
The AI gets a new ship every 200 AIP. That means there's only a 50/50 chance the AI will unlock another bonus ship type in a given game before you are attacking the AI homeworlds. That tells me that somewhere in the mid to mid-late game it is expected that you will hit 200 AIP so the AI gets a new ship.
New AI ships are based on unmodified AIP, so it is pretty much impossible to avoid getting them.  Taking the 8 planets needed to knock of CSGs put you at 170.  Gate raiding your home system, taking an extra planet, AIP over time, and the various AI toys that cost you AIP are all easy ways to pick up the last 30 AIP.

Offline chemical_art

  • Core Member Mark IV
  • *****
  • Posts: 3,952
  • Fabulous
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #32 on: April 12, 2013, 09:33:09 am »

Low-AIP games just circumvent those mechanics completly. From those values you'd think that "early game" (initial expansion) would be below 200, mid-game (preparing first assault) would be around 400, first assault up to 600, final assault around 800.

What is interesting I find is that in very late 2010, with me knowing nothing of the game, I thought these the values. It was thought for a time that as you long you didn't hit tech III units, the game would be winnable. Looking back, I find it outstanding that now even 300 AIP is considered "high" rather then mid.
Life is short. Have fun.

Offline chemical_art

  • Core Member Mark IV
  • *****
  • Posts: 3,952
  • Fabulous
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #33 on: April 12, 2013, 09:35:36 am »


Forgive me if I'm missing something obvious, but doesn't this suggest that the best fix would be to reduce the amount of knowledge available at the start of the game? This makes all game types more difficult, but hits low-AIP proportionally harder. Each extra planet taken is still worth the same amount to the player; you just start off weaker and therefore have further to climb.

This actually hurts higher AIP games more. This is because with c K, which increases the player's power is outstripped by the AI except for the very first few planets, if that. Reducing K makes the player's increase of power even less, favoring low AIP games more.
Life is short. Have fun.

Offline TechSY730

  • Core Member Mark V
  • *****
  • Posts: 4,570
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #34 on: April 12, 2013, 10:00:46 am »


Forgive me if I'm missing something obvious, but doesn't this suggest that the best fix would be to reduce the amount of knowledge available at the start of the game? This makes all game types more difficult, but hits low-AIP proportionally harder. Each extra planet taken is still worth the same amount to the player; you just start off weaker and therefore have further to climb.

This actually hurts higher AIP games more. This is because with c K, which increases the player's power is outstripped by the AI except for the very first few planets, if that. Reducing K makes the player's increase of power even less, favoring low AIP games more.

Which is why I guess you would also make the AI grow more slowly over AIP in return.

Which brings us pretty much right back to the original post.

Offline keith.lamothe

  • Arcen Games Staff
  • Arcen Staff
  • Zenith Council Member Mark III
  • *****
  • Posts: 19,505
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #35 on: April 12, 2013, 10:38:40 am »
Low-AIP games just circumvent those mechanics completly. From those values you'd think that "early game" (initial expansion) would be below 200, mid-game (preparing first assault) would be around 400, first assault up to 600, final assault around 800.
I don't think the expected values have ever been that high, at least not since January 2010 when I started helping with AIW.

In general the idea was "be very careful to not cross the line to mkII waves until you're ready, and if you cross the line to mkIII waves you're probably going to lose".

The extra bonus ships gained by the AI are mostly gravy to throw a wrench in, and it does get at least the 200-tier one in most games (that I'm aware of).

Quote
What I've seen during last years are lots of updates to make the AI stronger by adding mechanics to them, shouldn't we make the AI stronger by using mechanics that exists already ?
There have been a lot of changes in the last nearly-4-years (I wasn't with Arcen at the time, but iirc 1.0 of AIW came out in May '09).  Even within the last year, there's a pretty big metric ton of them.  So I don't expect you to be familiar with all of them.  But FWIW while we have added a lot of new mechanics (both on the AI and the human side), we've also strengthened most (maybe all) of the AI's existing mechanics too.  CPAs, waves, reinforcements, guard posts, etc.
Have ideas or bug reports for one of our games? Mantis for Suggestions and Bug Reports. Thanks for helping to make our games better!

Offline keith.lamothe

  • Arcen Games Staff
  • Arcen Staff
  • Zenith Council Member Mark III
  • *****
  • Posts: 19,505
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #36 on: April 13, 2013, 10:12:38 pm »
Ok, I've thought about this for a while and while I know the thread was a thought experiment to begin with I think the basic idea is worth trying, so here's what I'm thinking of doing (near-term; if it's well received this could be in 6.019, though I suspect there will be opposition, which is fine) for ratcheting back the "arms race" a bit to bring the expected AIP numbers (and by extension planets-taken count) back closer to how they used to be:


1) Reduce the effect of AIP to 2/3rds of current in the following uses:
- When determining the size of waves (including counterattack waves, etc).
- When determining the size of reinforcements.
- When determining the size of a CPA.
- When determining the rate at which the Special Forces grow.
- On Lazy-AI, when determining the rate at which Strategic Reserves grow (still capped at 200; Non-Lazy-AI would stay constant at 200).
- When determining the size of a Core CPA Guard Post's response to being triggered.
- When determining the frequency with which Broken-Golems-Hard/Botnet-Golem-Hard/Spirecraft-Hard exos come (there's a time-based floor to this that would remain unaffected, though).

These places where AIP is used would remain unaffected (this combined with the above is, to my knowledge from just now searching the code, a comprehensive list of things that AIP influences) :
- When determining the AI's tech level for ships in waves, etc.
- When determining when the AI should get new bonus ship type unlocks.
- When determining whether the human has too advanced a champion at too low an AIP and thus the nemesis response needs to ratchet up (so this remaining the same favors the player)
- When determining the Dyson's "population cap" for dysons in the galaxy.
- When determining how many Roaming Enclaves to spawn per event.


2) Reduce the max knowledge a player can gain from a planet from 3000 back to the old value (pre-4.0) of 2000.


The main caveat I can think of is that Fallen Spire would be nerfed somewhat by this as you'd probably need to take more planets to get the K you need to upgrade your FS stuff, and the FS exos would not be impacted by the above (not being based on AIP).  But I don't think it'd be a very big nerf because taking those extra planets doesn't make FS exos bigger either, and honestly once you have an FS fleet the AI's not exactly going to be able to stop you from taking those extra planets or punish you very much for the extra AIP they represent.


Anyway, thoughts?
Have ideas or bug reports for one of our games? Mantis for Suggestions and Bug Reports. Thanks for helping to make our games better!

Offline Hearteater

  • Core Member
  • *****
  • Posts: 2,334
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #37 on: April 13, 2013, 10:54:03 pm »
Without playing with it, my initial thought is, this won't make me take any additional systems, and the AI will be a relative cake walk at 9/9.  I say this because I have enough spare K already that the -1K/system just means I have less of an ending buffer.  Really the question I don't have an answer for is can I break the homeworld with my current strat given a static 200 AIP reserve.  I think I can, so this just makes the AI easier.

I don't have a problem with these change overall, I just think some additional adjustments will be needed.  Primarily in the power players gain with additional planets after about 10 systems.  I think Knowledge unlock value and energy are two areas this could be changed.

Offline keith.lamothe

  • Arcen Games Staff
  • Arcen Staff
  • Zenith Council Member Mark III
  • *****
  • Posts: 19,505
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #38 on: April 13, 2013, 11:38:14 pm »
Yea, potentially the energy collector would need to go from 200k to 150k to go with this, since it didn't exist in the 2k-K-per-planet days and wasn't balanced for it.  M+C I think can just be left as-is with this change, it would just theoretically make harvester upgrades less essential (not that I think it will stop people from just unlocking harvester IIIs immediately anyway, but oh well).

Then there's the question of knowledge costs in general.
Have ideas or bug reports for one of our games? Mantis for Suggestions and Bug Reports. Thanks for helping to make our games better!

Offline Hearteater

  • Core Member
  • *****
  • Posts: 2,334
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #39 on: April 14, 2013, 12:01:53 am »
Actually, I think Matter Converters are a problem.  Currently, I don't really need to make any choices with regards to energy.  Since I can trade M+C for Energy, and since I have effectively infinite time, I can trade M+C income down to very low levels and support max cap of everything I have.  It takes forever to rebuild, but for maximum power I'm encouraged to do that.  Worse, I can micro a refleet by destroying a chunk of Matter Converters to up my refleet speed until I hit my energy ceiling.

Another observation is each planet produces K Knowledge.  On average each point of Knowledge unlocks something that requires X Energy to support.  Each Energy Collector produces E Energy.  If E / K > X then Energy is largely pointless.  If E / K < X then Knowledge is largely pointless.  Basically, Knowledge and Energy are currently the same.  You get a fixed amount of each per planet (ignore Matter Converters because they just make Energy completely irrelevant).

I think ultimately something needs to happen to make them different.  Given the important of Force Fields to the human, and the current effect lose of Energy has on Force Fields, I'm inclined to think having Force Field strength scale on unspent Energy might work.  Matter Converters would need to go, and the Energy Cost of most units would have to drop so you could never spend even 50% of your energy (since you'd need a lot of surplus to get good Force Field strength).  There are a lot of other interesting balance features this presents, but since it's late and I'm sick so I'll leave off here.

Offline TechSY730

  • Core Member Mark V
  • *****
  • Posts: 4,570
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #40 on: April 14, 2013, 12:42:04 am »
Ok, I've thought about this for a while and while I know the thread was a thought experiment to begin with I think the basic idea is worth trying, so here's what I'm thinking of doing (near-term; if it's well received this could be in 6.019, though I suspect there will be opposition, which is fine) for ratcheting back the "arms race" a bit to bring the expected AIP numbers (and by extension planets-taken count) back closer to how they used to be:

(long snip)

Anyway, thoughts?

Looks intriguing, though both energy production values (both converter and collector) and starting knowledge (probably should come down some) would need some adjustment to have the intended effect of "slowing down" the arms race without making the AI a cakewalk.

Offline LordSloth

  • Sr. Member Mark III
  • ****
  • Posts: 430
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #41 on: April 14, 2013, 03:27:35 am »
Keeping along the lines of 'thought exercise', what if energy collectors only produced 100,000 energy per system, and economic command stations recieved a significant buff to energy production (roughly 50k,100k,150k)?

A). Military and Logistics command users would be seeing their energy production halved compared to current levels, and roughly a third below the proposed target. In theory military and logistics would compensate by increasing your firepower or at least the availability of your mobile forces, letting you get away with less.
B). Users of regular economic command stations would fall directly into the target energy profile, and wouldn't see any significant sacrifice from putting a few up on the front lines.
C). Users of mk2 eco commands would effectively have 300,000 free energy for spirecraft and/or golems.
D). Users of mk3 eco commands would effectively have 600,000+300,000 energy free for a botnet golem.
E). Low AIP Users of Warp Jammer Commands would quite possibly call for a bloody coup.
F). What kind of fortress energy costs would Cinth be comfortable with?

In the case of E) you might consider promoting Zenith Power Generators to objective level importance in terms of map placement, defensibility, etc - I'm not too familiar with low AIP games or 9/9.

If the starting level of power for a low AIP, high difficulty AI game needed tweaking, there are some available structures for it, some of which are constant (command stations) and some of which scale in number with difficulty (those human settlement-type things).

If you were spamming logistics commands:
At two planets grabbed, you'd have 150k/300k (target, current) less energy to work with before dipping into converters.
At five planets grabbed, a plausible 7/7 everything near your homeworld, you'd have 300k/600k energy less to work with.
At ten planets grabbed (currently less than attractive barring fallen spire, but more attractive if you factor in the K and AIP reduction, quite reasonable), you'd have 500k/1000K less energy to work with, but you'd also have less critical planets you could task to economic comands. This ten planets falls more inline with the wiki's suggestion for an opening start. On the other hand, if you took ten planets when you'd normally take six, then that 300k energy shortage would be reduced to only 100k energy shortage.

Rather than adapt to the existing rules of the game, I'm still heavily influence by my first impressions (and my co-op games) which were (excepting co-op) inspired by consulting the wiki for some idea of what my open game should look like. So far I've completed (and also abandoned) considerably more co-op games than SP, but I'm currently going through a batch of SP training sessions in order to provide some more current feedback, especially now that you're not competing with Dishonored and XCom and FTL, and before Age of Wonders 3 and a -lot- of kickstarter turn-based strategy games and Dragon Commander divert my gameplay hours temporarily. Generally speaking, I like your proposed suggestion Keith Lamonthe, and from the perspective of your average mediocre 7/7 greedy player I feel this can only be a good thing. That said, I've seen a bunch of interesting proposals that dealt with refleeting times (the whole AI is having a 'who is winning' meter, based on what you've done/lost lately that goes fully up and down being my favorite with the player just setting things up differently in the lobby being my second favorite).

http://www.arcengames.com/mediawiki/index.php?title=AI_War_-_Rate_Of_Expansion
http://www.arcengames.com/mediawiki/index.php?title=AI_War_-_AI_Progress#When_To_Worry

Considering the proposed excise of crystal in favor of a new improved hacking mechanic, my mind kind of hurts on figuring out the balance of converters and all that. Because on top of that, if eco commands got the proposed power boost, then they'd basically be worth two converters or more per system, if a person was willing to give up the extra resources.

Offline Faulty Logic

  • Hero Member Mark III
  • *****
  • Posts: 1,194
  • Bane of the AI
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #42 on: April 14, 2013, 04:02:37 am »
I'm ambivalent about this. I give it the much-cherished Tentative Approval prize.

I would say to reduce the energy output the energy collector to 100k energy, while allowing home to support 2 energy collectors or the equivalent.

I'm assuming k hacking would still generate antagonism per k, and that archives would now go only to 6000.

K costs do need some rebalancing, but not tonnes.

Quote
Looks intriguing, though both energy production values (both converter and collector) and starting knowledge (probably should come down some) would need some adjustment to have the intended effect of "slowing down" the arms race without making the AI a cakewalk.
Why would starting k need to come down? Each k point is still worth the same, it is just harder to get.
If warheads can't solve it, use more warheads.

Offline Fluffiest

  • Full Member
  • ***
  • Posts: 140
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #43 on: April 14, 2013, 08:06:08 am »
Ok, I've thought about this for a while and while I know the thread was a thought experiment to begin with I think the basic idea is worth trying, so here's what I'm thinking of doing (near-term; if it's well received this could be in 6.019, though I suspect there will be opposition, which is fine) for ratcheting back the "arms race" a bit to bring the expected AIP numbers (and by extension planets-taken count) back closer to how they used to be:


1) Reduce the effect of AIP to 2/3rds of current in the following uses:
- When determining the size of waves (including counterattack waves, etc).
- When determining the size of reinforcements.
- When determining the size of a CPA.
- When determining the rate at which the Special Forces grow.
- On Lazy-AI, when determining the rate at which Strategic Reserves grow (still capped at 200; Non-Lazy-AI would stay constant at 200).
- When determining the size of a Core CPA Guard Post's response to being triggered.
- When determining the frequency with which Broken-Golems-Hard/Botnet-Golem-Hard/Spirecraft-Hard exos come (there's a time-based floor to this that would remain unaffected, though).

These places where AIP is used would remain unaffected (this combined with the above is, to my knowledge from just now searching the code, a comprehensive list of things that AIP influences) :
- When determining the AI's tech level for ships in waves, etc.
- When determining when the AI should get new bonus ship type unlocks.
- When determining whether the human has too advanced a champion at too low an AIP and thus the nemesis response needs to ratchet up (so this remaining the same favors the player)
- When determining the Dyson's "population cap" for dysons in the galaxy.
- When determining how many Roaming Enclaves to spawn per event.


2) Reduce the max knowledge a player can gain from a planet from 3000 back to the old value (pre-4.0) of 2000.


The main caveat I can think of is that Fallen Spire would be nerfed somewhat by this as you'd probably need to take more planets to get the K you need to upgrade your FS stuff, and the FS exos would not be impacted by the above (not being based on AIP).  But I don't think it'd be a very big nerf because taking those extra planets doesn't make FS exos bigger either, and honestly once you have an FS fleet the AI's not exactly going to be able to stop you from taking those extra planets or punish you very much for the extra AIP they represent.

If you reduce the knowledge per planet, you're making each planet less valuable to the player, and encouraging the player to take fewer planets. This favours the low-AIP strategy further: If the K-per-planet is reduced by 1000, a player who takes three planets is 3000 K down. A player who takes a dozen planets is 12000 K down. Please consider reducing the player's starting knowledge instead.

It seems to me that in order to favour a high-AIP approach, you should raise the AI's starting strength, lower AIP scaling, lower the player's starting strength, and/or increase the benefit each planet gives to the player. To favour a low-AIP approach, do the opposite.

This, of course, comes with the caveat that making the AI's attacks too strong and the player too weak at the start of the game will result in a lot of losses in the first half-hour. It's possible that the AI's offense and defense should follow slightly different curves.

Offline Faulty Logic

  • Hero Member Mark III
  • *****
  • Posts: 1,194
  • Bane of the AI
Re: (Thought Exercise) Reduce knowledge to reduce AIP scaling.
« Reply #44 on: April 14, 2013, 08:16:56 am »
The purpose was that for the same amount of AI aggression, you get the same amount of k, but more planets, and thus either better map position or more capturables.
If warheads can't solve it, use more warheads.