View Issue Details
ID | Project | Category | Date Submitted | Last Update | |
---|---|---|---|---|---|
0024273 | AI War 2 | Balance Issue | Dec 26, 2020 11:27 am | Dec 29, 2020 11:24 am | |
Reporter | TechSY730 | Assigned To | Chris_McElligottPark | ||
Status | resolved | Resolution | fixed | ||
Product Version | 2.709 Obedient Engineers | ||||
Fixed in Version | Beta 2.710 The AIs Visit An Accountant | ||||
Summary | 0024273: AI is a bit too anemic at AIP 10 | ||||
Description | While the AI is supposed to be rather dismissive of you at game start, right now it seems like it is a bit too dismissive at AIP 10, especially at higher difficulties. This leads to the optimal strategy to not take any planets until you get some AIP reductions sniped. Due to just how beneficial it is to keep the AI so weak, and that there is no pressure at AIP 10 to get more resources. I propose _one_ of these two: * Make starting AIP goes up some (possibly only for higher difficulties) * Have a base (ideally moddable) budget term added to the budget calculations (so there is some budget even if the AIP hit a theoretical 0) If either is taken, a reduction on how much budget the AI gets per AIP may be warranted (but keep it a strictly non-decreasing relationship over difficulty of course, we never want an AIP where a higher difficulty has less resources). Credits to several players on the Discord | ||||
Tags | No tags attached. | ||||
related to | 0024272 | resolved | Chris_McElligottPark | AI budgets accelerates too quickly at low AIPs (especially higher difficulties) |
|
Hm, I believe the 50% AIP floor on diff 10 is quite drastic and complicates the game a lot, since it makes any reduction sort of small-ish, and the 10-AI gains massive strength from additional progress. As for the start, I always eventually run out of objectives, since scouting planets for hacks only gets you as far before spending so many hacking points that the hackable objects can no longer be used. And the AI really isn't much at risk from the player starting fleet, so no serious response seems fitting. Can you elaborate to which point your 10-AI games usually get? If the player cannot win, I personally don't see any problem letting him fly under the radar a couple hours if so he desires. |
|
He's referring to when the AIP is at 10, not the AI difficulty level 10. A slightly different situation. |
|
Thanks! * Added a new aip_absolute_floor for each AI difficulty level, with the following values: ** Diff 1-4: 10. ** Diff 5-6: 20. ** Diff 7: 30. ** Diff 8: 40. ** Diff 9: 50. ** Diff 10: 60. ** Note that the underlying starting AIP is still always just 10. But the effective AIP will never actually BE 10 on anything higher than difficulty 4 now, which basically has two effects: *** Firstly, it gets the game moving more quickly, with the AI being a lot less anemic at the start as the difficulty levels go up. The AI is increasingly quick to recognize you as you get up there in difficulty levels. *** Secondly, it does also give you a certain amount of "free AIP that doesn't change the situation" in each difficulty level right at the start. For instance, on difficulty 10, there are now 50 AIP that you can incur -- without any reduction -- before you even START to see the AIP rise from its new starting value. **** It's worth pointing out that this is not TRULY "free" AIP, because the floor percentage is still going up no matter what, and the game is already starting you out in a harder situation now, but it is essentially "free for a while," anyway. In the long form of a campaign it is not free at all, but in terms of short-term consequences in the first 15-20 minutes of a campaign, there's no change to the AIP responsiveness levels when you take your early objectives. ***** This means the early game is likely to be more difficult at higher difficulties in general, but there's also no need for strategies where you keep the AIP ultra-low (since you can't), and instead you have the flexibility of HOW to spend your first "free for now" AIP, and after that shift into thinking about how not to raise it any more if that's your strategy. ** The in-game tooltips have been updated to include wording about the new absolute floors that were added, including a special note if your actual AIP is currently lower than the absolute floor. ** The reason that we didn't just flat make the starting AIP higher (which would have in some ways been simpler) is that it would have been just a straight nerf to players, versus giving the higher difficulty levels this "free for now" grace period to play around in. Once you are to the mid-game on any difficulty level, this new "absolute floor" really is not likely to be a factor at all. This just gets games off the ground faster, and makes the AI progressively more dangerous in the early game on higher difficulties. |
|
This also beefs it up by 300% at AIP 10: * Removed from the ExternalConstants table: ** balance_seconds_times_planets_worth_of_aip_required_to_accumulate_one_cap="50" ** balance_ai_purchase_cost_per_cap_ai_income="400" ** balance_strength_per_cap_general="200" *** This has been replaced by balance_base_difficulty_for_raid_traversal="10", which is the equivalent of what was being used in the code previously. *** Arguably this shouldbe 20 instead of 10 in order for this to match what the AI budgets were based on in the past, but honestly the two concepts are not actually related. * Updated the math for GetOverallAIPurchaseCostGainPerSecond() ** The old logic was was effectively the following: *** EffectiveAIP / 20 *** Mult by balance_ai_purchase_cost_per_cap_ai_income (400) *** Mult by ai_purchase_cost_income_multiplier for the AI Difficulty (9.5 on diff 10, 2 on diff 7, 0.8 on diff 5). *** Divide by balance_seconds_times_planets_worth_of_aip_required_to_accumulate_one_cap (50). *** Then do some other stuff that is fine, and we'll ignore that other stuff for now. ** As you can see with the old logic on higher difficulties in particular, any and every AIP change upwards is incredibly painful, partially since there's no baseline income. Everything is added income completely dependent on the AIP. ** There's a spreadsheet with the new logic: https://docs.google.com/spreadsheets/d/1oSsxmKXDxP28Hc-ZYXngCnqnu3mc0vXipGEdj3HqKUo/edit?usp=sharing ** The new logic requires the following data changes: *** ai_purchase_cost_income_multiplier has been removed from AI difficulty levels. *** budget_base_income, budget_income_multiplier_for_AIP_Over_10, budget_income_multiplier_for_AIP_Over_50, budget_income_multiplier_for_AIP_Over_100, budget_income_multiplier_for_AIP_Over_200, budget_income_multiplier_for_AIP_Over_400, and budget_income_multiplier_for_AIP_Over_600 have all been added. ** The formula for the new logic is not perfect in terms of isolating each coefficient to each range, but it does a really good job of letting us tune the ranges in general and arrive at a variety of different slopes: *** BaseIncome+((MIN(50,CurrentAIP-10))*Mult_Over_10)+(MAX(0,MIN(CurrentAIP-50,100))*Mult_Over_50)+(MAX(0,MIN(CurrentAIP-100,200))*Mult_Over_100)+(MAX(0,MIN(CurrentAIP-200,400))*Mult_Over_200)+(MAX(0,MIN(CurrentAIP-400,600))*Mult_Over_400)+(MAX(0,CurrentAIP-600)*Mult_Over_600) *** The end result is something that is substantially more income at the lower difficulties (4 and down), while still keeping them very tame. They are not so incredibly poor as before. *** At AIP 10, the general bump is at least around 300% over what it used to be, but on difficulty 5 the budget at AIP 50 is only 8% what it was. For difficulties above 5, the budgets are only about 75% of previous at AIP 50. *** Overall the budgets are higher for AIP 4 and down, and in general you can see how that works on the spreadsheet here: https://docs.google.com/spreadsheets/d/1oSsxmKXDxP28Hc-ZYXngCnqnu3mc0vXipGEdj3HqKUo/edit#gid=1751215831 |
Date Modified | Username | Field | Change |
---|---|---|---|
Dec 26, 2020 11:27 am | TechSY730 | New Issue | |
Dec 26, 2020 11:28 am | TechSY730 | Relationship added | related to 0024272 |
Dec 26, 2020 11:42 am | TechSY730 | Description Updated | |
Dec 26, 2020 11:42 am | TechSY730 | Description Updated | |
Dec 26, 2020 11:44 am | TechSY730 | Description Updated | |
Dec 26, 2020 4:02 pm | TechSY730 | Description Updated | |
Dec 26, 2020 4:04 pm | TechSY730 | Description Updated | |
Dec 28, 2020 5:13 am | Arides | Note Added: 0060163 | |
Dec 28, 2020 2:44 pm | Chris_McElligottPark | Note Added: 0060165 | |
Dec 28, 2020 3:06 pm | Chris_McElligottPark | Assigned To | => Chris_McElligottPark |
Dec 28, 2020 3:06 pm | Chris_McElligottPark | Status | new => resolved |
Dec 28, 2020 3:06 pm | Chris_McElligottPark | Resolution | open => fixed |
Dec 28, 2020 3:06 pm | Chris_McElligottPark | Fixed in Version | => Beta 2.710 The AIs Visit An Accountant |
Dec 28, 2020 3:06 pm | Chris_McElligottPark | Note Added: 0060166 | |
Dec 29, 2020 11:24 am | Chris_McElligottPark | Note Added: 0060173 |