View Issue Details
| ID | Project | Category | Date Submitted | Last Update | |
|---|---|---|---|---|---|
| 0024677 | AI War 2 | Balance Issue | Apr 12, 2021 12:28 pm | Apr 22, 2021 1:39 am | |
| Reporter | Metrekec | Assigned To | |||
| Status | new | Resolution | open | ||
| Product Version | Beta 2.773 Civvies Hotfix | ||||
| Summary | 0024677: Double Supercat AI died Beta 2.773 | ||||
| Description | Reporting this as ArnaudB suggested. Based on https://forums.arcengames.com/ai-war-ii/standards-for-pure-10-or-1010-wins-vs-any-10-or-1010-wins/ this follows the correct guidelines for "Reporting the bug". Two Full ensemble 10/10/10. Base game + DLC1, Vanilla, default options. Realistic map, 80 planets. Mostly played at 3x speed, though did use 5x speed occasionally while waiting for construction/metal. Demolition starting fleet with Raider defenses and Divorce defenses. Splash had been very effective as a build in the past for me, so I figured I would give it a shot. Divorce defense battlestation was hardly relevant and most of my beachheading was done with Raider, with some translocation turrets and spider turrets occasionally thrown in (mostly used as a CC battlestation). The agile transport was also very helpful since the extremely fast ship helped me reposition my fleet and confuse the warden/hunter about which planets were "safe" to attack. Main techs were Splash (3) and Raid (2), with the majority of my strikecraft going Splash and the majority of frigates being Raid Frigates. Once again, I feel that the turrets were the star of the show, particularly for the homeworld assaults, but with some proper micro both Raid and Splash could be used effectively to take out key targets and large swathes of AI ships respectively. The buffs to corvettes was another reason for picking Splash. I did not mess around much with a splash focused build in the past, but considering the amount of ships I was up against, I figured I'd give it a shot. They were very effective and I'll definitely be using them again in the future. TTS and ARS hacks were similar to my last 10 win. 3rd hack is still prohibitively expensive. Likely due to reduced budget on both AIs during the early game, I delayed getting a second turret hack for a long time and was able to defend adequately with only 1 hack and some battlestation turrets. I got a lot luckier on my ARS choices this time around, so I picked up more strikecraft and frigates than I did during my solo 10 win. Initial start had 2 transports, TTS and ARS, one support fleet and one Ark (Ark One) in my 2 planet cubbyhole, totalling (after 2 ARS hack) 4 assorted Mk1 strikecraft, 4 splash strikecraft, 1 splash frigate, 1 raider strikecraft, 2 raider frigates. These types of starts are extremely helpful in 10 difficulty as the early-game is when the AI's budget feels its most insurmountable. Without the initial power boost of a good start, 10 would be a lot harder. However, I feel like 2 transports with decent techs would still be sufficient to manage the early-game with proper battlestation usage, so I don't think any balance here would do much more than make the early-game grindier. Later on I also picked up a thanatos ark as it was in my path, and it ended up doing nicely, but I don't think the arks were extremely relevant. They definitely help out, but they aren't the star of the show or anything. Both were marked up to 4 for most of the game. Starting with the raid battlestation was surprisingly good for defense. Makeshift turrets and Blitz turret's drone spawns mean that they are infinite range turrets that also have good close range damage, and the effectiveness really impressed me. I felt these turrets carried me defensively through the early and the midgame, possibly they are currently overpowered. Ruffian and Scrap turrets were mostly used for early-mid beachheads. Did not need to put as many turrets early on in my Military Station choke due to the reduced budget of both AIs again. 2 AIs at 75% budget do not coordinate, so even if it may seem like 150% budget at the end of the day, as long as the budget is scattered, it is not a big deal. Since I did not need as many turrets to defend myself and econ energy amounts have been buffed, I only ended up using 1 military + econ rather than 2 during my last 10 win, supporting my other military command stations which were both placed on MDCs. There were some scary situations with the MDCs as one in particular was placed very close to a collection of wormholes, but as I made sure to put maximum turrets on both, the AI mostly concentrated its attacks on the "weaker" home command station chokepoint which was less strength overall but still plenty strong to deal with whatever came its way (Once again, 75% budget helps). Speaking of defenses, the forcefield nerf since my last 10 win did not feel that impactful. In most cases my turrets were strong enough to kill the AI before too many scary ships reached the forcefields, and the ones that did were still not strong enough to threaten me too often. There were a couple of cases where I had to micro the forcefields and engineers to make sure that the stations survived, but overall I think they are still quite good even without significant science investment. I ended up putting 2 points in them throughout the game, and that felt plenty. As logistics lost their focused gravity generators, I now hack them from the ODSS whenever I see them. This gives military and logistics command stations 6 (!!!) and my battlestations 4 (!) focused gravity generators which is probably the biggest increase to my defensive capabilities for its cost. They also help out in beachheads as they are cheap to build and do not tend to die quickly, slowing down the AI while I build up the rest of my turrets. These amounts, for 20 HaP, could probably be halved and still be worth getting in my opinion, though other players may disagree. Personally I feel that this is the most important part of my mid-lategame defenses. Right now if an ODSS does not initially roll gravity generators, I will pay 20 HaP on another pickup that's less relevant (usually interplanetary engineers) in order to get the gravity generators, as 40 HaP still feels like a good price for them considering how versatile they are. Logistics to Military baiting is back, and was used a few times in the midgame to bait hunter build-up. I think this is cheesy and should have bigger penalties, and was a fan of when forcefields were forcibly destroyed by the swap due to both station types having different forcefields. I understand this was removed because it was frustrating, but I think a similarly punishing penalty should be added, or swapping should be disabled completely while a planet is under attack. The AI cannot properly react to the threat of a player suddenly switching a weak logistic planet to a military after committing a bunch of forces. Once again, every command station got 500 science upgrade when built. The benefits of instantly investing 500 science feel too good to pass up on, as it impacts everything - defenses, economy, energy. Praetorian Guard was not as challenging this time around as they did not pre-emptively camp wormholes, so I was always able to beachhead before they reached me. They were still quite a pain though and the first encounter with them was tough as in my last diff 10 win. They were particularly relevant this game early on as they were protecting an MDC and the planet I used for my military+econ setup. Speaking of MDCs, I got both relatively quickly and was therefore set AIP-wise for the rest of the game. I still ended up hitting any data center I found along the way (which ended up being quite a few of them), but I was at the AIP floor during the entire match. Maybe MDCs give too much reduction and could give less reduction per MDC, as two -60 is huge and the bonus hunter fleet ships have never felt impactful enough to matter to me. The final key to this game, however, was the endgame. I've reached this state in the past against double 10 and found it very hard to start the AI bastion worlds - this is when the 150% budget starts to become seriously relevant, as the AIP starts to cascade. Because of this, I decided to trigger a civil war and quickly kill one of the AI's while both AI are suddenly thrown into anarchy. This resulted in a relatively easy first AI kill - not only did I have very strong strikecraft and beachhead from double spire archive hacks and extra planets due to all this AIP reduction, their warden and hunter were busy dealing with the other AI - same as their waves. The Overlord was also not too bad and I managed to kill it despite not having turrets specialized for killing big things - with a citadel you can build so many that the moment the overlord passes through a military world, it's toast. What I did not expect was how the 2nd AI would go. I was a bit worried as I knew it would go back to 100% budget and have whatever remained of the civil war leftover and roaming around. It would have extragalactic ships, the PG would be extremely strong, and any beachheads would be extremely difficult to set up as its forces would be all over its bastion worlds. My initial plan was to use my raid frigates to slowly and painstakingly take out the dire guardposts on each bastion planet before triggering the overlord, while dealing with whatever ended up generated with bait tactics, eventually triggering phase 2 and killing it in the same way I killed the first AI. However, I did not anticipate the remnants of the first AI's civil war hunter fleet - around 4-5k strength of it, suddenly barging down the second AI's door and completely sweeping it. This hunter fleet eventually ended up back at my home command chokepoint as well, presumably deciding that no one deserved to live despite being painted in "NOT AFTER YOU", but my defenses (and hastily recalled fleet) were able to grind it all out. Being able to respec science out of a few of my no longer needed command posts and boost my chokepoint to ludicrous strength levels meant that their attack ended up going nowhere. After that, it was just a matter of waiting for the 2nd phase 2 to reach the chokepoint and get killed as well. Overall it was very fun, though I personally felt the ending was a bit anti-climactic. I expected to have to put a lot more effort into finishing the 2nd AI, but it turned out that the 1st AI's rage after dying could not be contained! With how this game went, it almost felt like I was playing a "fair" difficulty, though. I don't know if 10 should feel like that, even with a strong start like I had. If that's how 10 is with the new game design, I have nothing against it. I really enjoyed it after all! As with last time, attaching a zip with all of my saves from the campaign. Last 20 minutes after the first AI kill don't have as many saves as I did not expect things to go as they did. I just watched in awe at how things ended up turning out! | ||||
| Tags | No tags attached. | ||||
|
|
|
|
|
|
|
|
Completely agree about the command station planetary upgrades. I think it's been long enough now that we can be pretty sure they're a little too optimal to be good for player choice. Right now they feel so strong you'd be mad not to use them for anything except fallen spire, and even there they're really good. Side suggestion: With the (relatively) high number of Diff 10 AI reports there seem to be, I can't help but wonder if perhaps diff 10 simply isn't enough of a step up from the 'normal' difficulties, and needs a bit of a raw budget boost comparatively? |
|
|
I think the first AI turning on the second AI and helping you win is a pretty epic ending. The two ideas that stod out most to me were: Cranking up the anti-MDC bonus hunters at diff 8+ I like the notion of preventing you from swapping command station types when there are enemies on the planet (or making it change type more slowly). |
|
|
Adding a new save which was won in 2.806 with the same principles except this time with Raid and Piercing. Piercing was main teched shipline-wise while Raid was main-teched turret wise. I had the usual good start (4 piercing ships + 1 raid ship on top of the raider starter fleet). I got lucky with a raid tech vault which saved me 5000 science, but otherwise the run was very similar to this raid/splash run. The biggest difference is that I won without triggering a civil war this time, properly finishing off the second AI myself. New AI Overlords (though I rolled 2x PFFN which was unfortunate - I wanted to get both variants in one swoop!) are very tough and I only barely managed to kill them when they assaulted me, requiring me to pull back everything and hastily construct at least 90% of my battlestation turrets on top of my usual chokehold. Their shield count is so high that it feels pretty pointless to engage them on anything except a highly leveled military planet - most other attempts to dent it simply gets absorbed by the shield before it escapes, that 500 speed is killer! Blitz turrets feel way too effective on defense to really be "raid" turrets. As in my main post, it and makeshift feel like they are way too powerful, and raid in general is a bit ridiculous in how quickly it builds excessive strength on AI planets, causing warden/praetorian to immediately rout and saving a lot of metal overall (as fights with the AI on 10 don't really benefit the player much since their budget is essentially bottomless, particularly at high AIP). New military scaling combined with this caused the threat buildup and waves at high AIP to be completely inconsequential. I had 2k threat roaming around, and even with all that roaming the hunter wasn't able to support waves at 300 AIP even combined with CPAs to mount attacks. See video: https://streamable.com/d33cfg I struggled pretty hard transitioning from early to midgame, but it's not clear if that was because I didn't have a great map RNG or if there were other factors. Once I got through a specific hump and reached my first spire archive, things went a lot smoother. To deal with the two AI's warden/hunter fleets ending up being overpowering during the first AI kill, I took two chokepoints in the middle of the map, preventing the wardens and hunters from both sides from reaching eachother. I'm not entirely certain how effective this ended up being, but it worked, so I'm glad I did it. Second AI was killed in the way I envisioned it in my first game. The massive blobs of praetorian were baited by going after turrets I placed around the gravity well while my fleet zipped around to kill the guardposts with the transport. I don't think there's much that can be done to prevent this barring better micromanagement AI from the praetorians - the transport is extremely fast and letting the turrets build up would mean that I kill the praetorian blob conventionally instead, and my turrets build up quite fast thanks to all my engineers and very high income this late in the game. As a side note, it seems like hunter that goes after the first dead AI eventually goes after the player once the first dead AI has nothing left - or at least, they started going after me at some point. I'm not sure if this is intended or not, but since my military stations were so powerful, it ended up not mattering much. |
|
|
|
|
|
|
|
|
This game was also very long - 2nd AI was killed at 9h 19 minutes (!) which is why the save dump is so large |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| Apr 12, 2021 12:28 pm | Metrekec | New Issue | |
| Apr 12, 2021 12:28 pm | Metrekec | File Added: Double10_p2.7z | |
| Apr 12, 2021 12:28 pm | Metrekec | File Added: Double10_p1.7z | |
| Apr 12, 2021 12:28 pm | Metrekec | File Added: Double10_p4.7z | |
| Apr 12, 2021 12:28 pm | Metrekec | File Added: Double10_p3.7z | |
| Apr 12, 2021 1:09 pm | Lord Of Nothing | Note Added: 0061093 | |
| Apr 13, 2021 2:34 am | BadgerBadger | Note Added: 0061102 | |
| Apr 22, 2021 1:37 am | Metrekec | File Added: Raid10x2_1.7z | |
| Apr 22, 2021 1:37 am | Metrekec | Note Added: 0061194 | |
| Apr 22, 2021 1:38 am | Metrekec | File Added: Raid10x2_2.7z | |
| Apr 22, 2021 1:38 am | Metrekec | File Added: Raid10x2_3.7z | |
| Apr 22, 2021 1:39 am | Metrekec | File Added: Raid10x2_4.7z | |
| Apr 22, 2021 1:39 am | Metrekec | Note Added: 0061195 |
