There are at least three different good points here.



a. Micromanagement. I am pretty sure there are already individual suggestion threads for all of these; please see jetkar's suggestion summary thread linked in my sig. A number of people have all complained of the same things.



b. No battle cards in auto resolve: I suspected this, but it is hard to "prove". Can anybody demonstrate for sure that neither side plays *any* cards in auto resolve? I thought that the game just picked "poorly" in auto resolve, and that is why manual resolve is better.



c. AI improvements. A lot of these points are discussed on my thread:

/#/endless-space/forum/27-general/thread/9509-evidence-requested-poor-ai-decisions-besides-ship-design

I will reply to the AI points there, to attempt to centralize the suggestions.

davea wrote:
I agree these are indications. But it is possible that full-auto resolution uses a different system entirely based on some MP ratios rather than the visual battle phases. And the heal card could just indicate that the part of the game code which chooses cards (if there is one) doesn't like this particular card. Can we construct an experiment to *prove* this?

I'm pretty sure that it isn't just based on total MP ratios or such - I use auto-battle literally all the time, and the AI often has considerably more MP than me, but my fleet still consistently wins unscratched due to having much better ship design (proper defenses).

So it does seem to run the same "simulation" as manual combat would do. It does seem to completely forgo cards for both sides, though.

Well with battles that you would be able to control, Homeworld style, it give the advanced player more to do and have more influence on the flow of the game.



I for once would LOVE to have the ability to use positioning and tactics to defeat a far superior foe with inferior ships, this mainly happens to me in the first 30 turns on Impossible or Endless, after that it becomes more about how many fleets AI can spit out.

Atlantis_Risen wrote:
Um, no. Managing my systems is why I play a 4X strategy game. The more complex the better. There's enough dumbed down games out there already.

It is not even true micromanagement, you just have to click very often and wait for the autoresolve to finish. If you want optimal Results you'd have to fight it out by hand every time, whih can become boring since we're dealing with mega stacks of fleets.

davea wrote:
I agree these are indications. But it is possible that full-auto resolution uses a different system entirely based on some MP ratios rather than the visual battle phases. And the heal card could just indicate that the part of the game code which chooses cards (if there is one) doesn't like this particular card. Can we construct an experiment to *prove* this?

I don't think auto being purely based on MP is plausible, I've had fleets take far less damage on auto without increasing their MP after I changed over to a better defense allocation. And given that the AI has been observed to choose the repair card at 100% health, it seems unlikely it is singled-out for non-usage.

Licho wrote:
Indications that auto resolve don't use any cards:

* in multiplayer, if one player auto resolves, and other playes manually, the auto -resolver fleet uses no cards

* in SP, auto resolve never heals ships

I agree these are indications. But it is possible that full-auto resolution uses a different system entirely based on some MP ratios rather than the visual battle phases. And the heal card could just indicate that the part of the game code which chooses cards (if there is one) doesn't like this particular card. Can we construct an experiment to *prove* this?

With the game in its current state it seems that if you manage to survive past turn 50-60 ( am playing on tiny small and medium galaxies with all 8 empires, tiny is really interesting using this setup) its pretty much a win form than on.



I am switching between Impossible and Endless difficulties and tying different trait combinations as well and am getting pretty consistent results as mentioned earlier. I am unsure how to solve it but for me it seems that the biggest issue is that the AI does not seem to adjust fast enough to the situation at hand.

davea:



Indications that auto resolve don't use any cards:

* in multiplayer, if one player auto resolves, and other playes manually, the auto -resolver fleet uses no cards

* in SP, auto resolve never heals ships



Atlantis_risen:



Notice the emphasis on *boring* micromanagement.

Micromanagement can be fun but when it turns into mindless clicking just to achieve something trivial and you have to click the same sequence over and over again it becomes boring.



Its fun to decide what to build.

Its not fun to push something to queue on all planets or to click the same sequence on all planets or to find idle planets

I am not sure that answers the question of what happens in a fully automatic resolution, where neither side has a GUI to view the action.

davea wrote:
b. No battle cards in auto resolve: I suspected this, but it is hard to "prove". Can anybody demonstrate for sure that neither side plays *any* cards in auto resolve? I thought that the game just picked "poorly" in auto resolve, and that is why manual resolve is better.

Easily proven:

Step 1)Start a 2 player multiplayer game(unique clusters)

Step 2) Players send scouts to middle of map to find each other and engage in battle.

Step 3) Player 1 selects Auto and Player 2 selects Manual.

Step 4) Player 2 Observes combat, may play battle cards or not as desired.

There's a difference between complex and tedious.

Licho wrote:
but full of boring micromanagement

Um, no. Managing my systems is why I play a 4X strategy game. The more complex the better. There's enough dumbed down games out there already.

I only choose auto - battle when then odds are steep in my favor. Otherwise if I know I am retreating I am forced to enter battle just to select that card or if my fleet is an attrition fleet I know would not win on points but wins because its bogus

I have always good terrific results from the Auto-battle, even close battles were within an acceptable margin of error!



I feel like the AI never has much of a meta plan for war, i mean your goal is always to destroy you opponent, but how do you do it?



If the AI could focus on shorter wars of smaller gain they might be able to better micromanage their systems, and then go back into a war they can manage. AI fleets should prioritize strategic resource systems above all else, even blockading them prevents their use and when done decisively can change wars.



The AI should possibly focus on more defensive fleets when entering into a war as well, so that they might be able to adapt without letting their first fleets be completely destroyed in the process.

First of all, the difficulty is not enough yet. Its still pretty trivial to beat even the hardest AI and its frustrating that the game uses maluses for player to increase difficulty.



Second the challenge isn't that much in thinking but in micromanaging like a robot basically:



1) managing queues on dozens of systems is cumbersome

- governors are too dumb compared to player

- common tasks like "insert structure X to all planets queue as first item" are tedious to do

- lack of some sort of custom player "piority queue" which would let you define prefered order of general improvements to follow on systems (or even template for governors)



2) managing planets forces you to check every system manually almost every turn

- you need to go to system details to check on your planets

- you need to go to individual planets to change specialization or to scout moon or remove anomaly - this makes turn very slow

- it would be nice to have automated "virtual-improvement" like "scout all moons" or "remove all anomalies" (with variable cost of course)



3) managing/merging fleets - there is no easy way to set rally points and so you have to go to hangars of lots of planets and order ships around every turn to send them to rally system



4) fighting battles on manual takes too much time

- there should be a way to define cards and skip the animation. Yes animations are great for new players but after spending hundreds of hours with the game you just set cards and leave for a toilet.



5) "idle systems" are hard to discover. They should use conver to dust or sci by default. Atm you can either go there from the news event or from empire list, but both suffer from memory loss. When you return from system detail you have to scroll down again.

What you said about micromanagement is spot on... but that is to be expected. AI improvements would help a lot.

Its what would happen if no one picked cards, and if the random number god wasn't a sadistic git!

Igncom1 wrote:
Nobody gets cards in Auto battles, its a clean battle.

Oh. Well that's new information, I just thought they used crappy cards because somehow 3 dreadnoughts lost against a couple of frigates at full health. -_-

Landswimmer wrote:


THIS! Oh my god yes. There should be automatic - Pick cards - and Manual. I hate losing some fleets because I'm lazy and I hit the auto button and the it picks the worst cards and I lose unneeded ships.

Nobody gets cards in Auto battles, its a clean battle.

The best way to program the computer is to get it to make the same choices we would make, system has 2 moons? its ok to build the 'clean sweep' improvement.



The more strategys and reports we get here can then be used to make an AI that acts like a player, so thanks for the support!

1) managing queues on dozens of systems is cumbersome

- governors are too dumb compared to player

- common tasks like "insert structure X to all planets queue as first item" are tedious to do

- lack of some sort of custom player "piority queue" which would let you define prefered order of general improvements to follow on systems (or even template for governors)



I agree



2) managing planets forces you to check every system manually almost every turn

- you need to go to system details to check on your planets

- you need to go to individual planets to change specialization or to scout moon or remove anomaly - this makes turn very slow

- it would be nice to have automated "virtual-improvement" like "scout all moons" or "remove all anomalies" (with variable cost of course)



I completely agree, this does need some fine tuning.





3) managing/merging fleets - there is no easy way to set rally points and so you have to go to hangars of lots of planets and order ships around every turn to send them to rally system



Rally points are needed



4) fighting battles on manual takes too much time

- there should be a way to define cards and skip the animation. Yes animations are great for new players but after spending hundreds of hours with the game you just set cards and leave for a toilet.



THIS! Oh my god yes. There should be automatic - Pick cards - and Manual. I hate losing some fleets because I'm lazy and I hit the auto button and the it picks the worst cards and I lose unneeded ships.



5) "idle systems" are hard to discover. They should use conver to dust or sci by default. Atm you can either go there from the news event or from empire list, but both suffer from memory loss. When you return from system detail you have to scroll down again.



Just like for science there should be an option to make the game remind you when there is nothing in production

Licho wrote:
things like food on moons are rarely worth it, clean sweep only with 2+ moons and AI can still pick them

The tooltip is misleading, those improvements actually give +3 Food or +2 Dust per population on the planet. So they are actually quite strong.

Renamed the title.

Some things that AI does in less optimal way:



Early colonization problems:

* colonizers - it should use colonizer with extra engine most of the time and expand more aggresively. With pilgrims it could use evacuate fleet and hero with +25production to expand.

* wormholes - basically your first goal is to get casimir and expand to "common" area to grab the best planets there. AI ignores that.

* spatial awarness - AI does not try to block key access points to rich system areas like humans do, thats probably hardest to implement, it could guard wormholes instead - even with simple scout. Rich planets too.

* hero switching - if rich system is identified close to plenty of empty planets it should assign its +25prod hero there and start colonizing from there

* initial planet setup - AI should focus on food on first planets initially and start queues with heavy isotopes



Tech tree problems:

* AI can be beaten by focusing on "economic" techs - which increase production or science. Things like +40 sci improvement are incredibly valuable early on to give you big bonuses

* AI appears to invest too much into warfare tree even before its needed. It should probably research that only if it plans/is at war or if it needs utility techs from there.



System improvements problems

* AI builds improvements that dont make sense or have only marginal returns. For example it can build influence or trade improvements on system where trade or influence dont make sense. Or defenses on planets which arent under attack.

* AI appears to try to specialize system. As a human who beats that AI i never do that. That probably indicates some problems with game design. What I do is focus on food first on all planets and when planets are near full i switch them to most efficient specialization for given planet type. At least until you have improvements like +30% +40% sci it makes no sense to specialize whole system.

* some improvements are almost never worth it yet AI can still build it - things like food on moons are rarely worth it, clean sweep only with 2+ moons and AI can still pick them.

* AI should insert into queue key great improvements when available - things like heavy isotopes or planetary institute. Basically if improvement can improve industry output and return time is short enough push that to queue first. Same with approval rating.



Diplomacy:

* the patch from yesterday is a step in the right direction, AI's now declare war on you when you grow too strong. However they should still do that earlier and swap target when someone else is strongest.





Also it appears that endless empire victory is too easy, because you have to explore that part of the tech tree anyway. You need the ship hulls, expansion disaproval fixes and planet improvements from the "Exploration and Expansion" tree so its usually very easy to finish the path by getting endless empire. What I also like to do is transforming lesser planets to desert (even arctic -> desert) and later to tundra.

Licho wrote:
3) managing/merging fleets - there is no easy way to set rally points and so you have to go to hangars of lots of planets and order ships around every turn to send them to rally system

I've been bothered by this myself, and it seems like it would be an easy one to fix.

Thanks for the feedback. Several of the suggestions you make are already under discussion since other players have suggested them too. Please look into the suggestion summary thread listed in my sig. In particular, I would love rally points and a way to preset cards for auto battle resolution.



Do you have any suggestions on what the AI should do, to be more of a challenge? We have collected a number of ideas here:



/#/endless-space/forum/27-general/thread/9509-evidence-requested-poor-ai-decisions-besides-ship-design

I came here from playing civilization 5 and this game while being inferior in some aspects to civ 5 (in my opinion) still pretty much the same game. However I do like the combat and diplomacy of this game a bit more than I did in civ 5. Anyway my point is - all civilization games that I have encountered require massive micro management. If you don't like it - pick smaller map.