Finally tried 0.4.8...

Describe your experience with the latest version of FreeOrion to help us improve it.

Moderators: Oberlus, Oberlus

Forum rules
Always mention the exact version of FreeOrion you are testing.

When reporting an issue regarding the AI, if possible provide the relevant AI log file and a save game file that demonstrates the issue.
Message
Author
ovarwa
Space Kraken
Posts: 106
Joined: Thu Dec 11, 2014 3:22 am

Re: Finally tried 0.4.8...

#61 Post by ovarwa » Thu Nov 15, 2018 7:39 pm

Hi,
Jaumito wrote:
Thu Nov 15, 2018 7:43 am
ovarwa wrote:
Thu Nov 15, 2018 3:08 am
Cluster galaxy: Weirdly, even though I have been using the same 300 stars and 12 players for nearly all of my games, I feel that there is a lot less pressure before conflict with the AI, and more room to expand. Don't know why.
Lots of chokepoints, that's why.
Makes sense, thanks!

Anyway,

Ken

ovarwa
Space Kraken
Posts: 106
Joined: Thu Dec 11, 2014 3:22 am

Re: Finally tried 0.4.8...

#62 Post by ovarwa » Thu Nov 15, 2018 8:45 pm

Hi,
Ophiuchus wrote:
Thu Nov 15, 2018 10:24 am
ovarwa wrote:
Wed Nov 14, 2018 7:00 pm
"Oh no, someone has been playing civ6."
Yes that kind of comment is pretty useless. On the other hand if you would have reasons why that is bad or how that could be improved maybe could make the whole thing better.
It's useless because I won't be able to have much of a conversation about it with someone who is a fan of civ6 policies after having seen what happens in civ6. It's even more useless than convincing my wife that beans are yummy; that, at least, is completely subjective.

Aside from that, I think it's fundamentally a bad idea because it dilutes whatever is emerging to make FO distinctive, in favor of making the game similar to the new hotness (which I don't think is so very hot, but that is a separate issue.) The slow development time of FO makes this kind of thing especially problematic, because there will be many new hotnesses that can be bolted onto the game, creating a mishmash of mismatched features.

How to do it better? Don't! The game has no need of governments at this time. Only one game with governments did an outstanding job, imo: Alpha Centauri. Only one game did a good job, imo: Civ4. The others (GalCivX, MOO2, CivX(X!=4), ...) are pretty much just there because CivOrig had (vestigial) governments, and governments make a lot of sense in that context.

They don't make very much sense in this context. In Civ, governments distinguish among human civilizations and how they develop. But when comparing alien species, they don't make sense. A human dictatorship and a human democracy are still fundamentally human, and aliens who are not specialists will have difficulty telling the difference; conversely, where Eassaw might see fundamental differences between those who want to kill aliens to death in the mines and those who want to kill aliens to death by eating them in a series of grand festivals, other species might not appreciate these critical distinctions.

But, you know, this kind of thinking is not useful in the face of "I want governments."

Nor is the thought that governments just make the AI problem worse because they are yet another bonus that only human players can really leverage (or they don't make the problem worse because the AI won't use them and they are so not worth it that humans won't use them either.)
ovarwa wrote:
Wed Nov 14, 2018 7:00 pm
If even a single brain cell that could be used for teaching the AI how to group fleets is instead caught up in influence, policies and governments, these features are problematic at this time. Or rebalancing the tech tree. Or fixing hulls.
I certainly disagree.. AI and rebalancing the tech tree are personal pain points (which are only relevant because the game is playable).
Of course you do! Most people here do.


For the 1.0 release ("feature complete") especially tech tree, species, hull rebalancing at this time is most probably a waste of time/effort. (Of course it would be great if someone would tackle it for 0.5 including influence and policies).
Especially for python/C++ developers (because we dont have enough of those).
Rebalancing effects is a huge effort and you basically have to redo it with every change/addition of major features. And you need a lot of playtesting to do it right (most FO developers do not have time for this). But you dont have to be a developer to fix that kind of problem.

I think c++ developers should improve in direction of 1.0 release ("feature complete", influence, fixing stealth mechanic), also python developers should work on improving AI mostly in combination with features which are probably not going to change.
I certainly disagree... I think improving toward a 1.0 release involves taking a ruthless look at features to see what can be ignored or implemented vestigially, and maybe even dropped, and doing a feature freeze.

But that's just me!

Anyway,

Ken

Jaumito
Space Kraken
Posts: 189
Joined: Tue May 16, 2017 3:42 am
Location: Catalonia, France, Europe, Earth, Sol, Orion Arm, Milky Way, Virgo Cluster

Re: Finally tried 0.4.8...

#63 Post by Jaumito » Fri Nov 16, 2018 12:16 am

ovarwa wrote:
Thu Nov 15, 2018 7:36 pm
What are your issues with fleet management?
Lots, but just to give a couple examples of its... unwieldiness:

I have a fleet of 100 ships on system A (let's assume they're the same type), and to react to a double threat I want to send the 70 less damaged ones to system B and the rest to system C. In a game like Stars!, I could do this in a handful of clicks. How many does it take in FO?

Let's assume now that these 100 ships are mostly Tits, with some support ships - let's say at least 3 each of facilitator (repairs), scattered (shields) and solar (fuel) hulls - now how do I organize all those ships in 3 different fleets with at least one of each type of support ship? Again, what was quick and easy with Stars! turns into a logistics nightmare in FO.
I do find that clicks and hand movements add up, so you might have noticed some I take for granted. (I think *you* are taking the stuff you consider 'nothing' for granted.
Hardly. What I'm saying is, while there are certainly places where you could divide the number of required mouse clicks by a factor of, say, about two (and that would indeed be an improvement), in the examples above I'm talking about a factor of 100 or more. It doesn't even compare.

ovarwa
Space Kraken
Posts: 106
Joined: Thu Dec 11, 2014 3:22 am

Re: Finally tried 0.4.8...

#64 Post by ovarwa » Fri Nov 16, 2018 12:32 am

Hi,
Jaumito wrote:
Fri Nov 16, 2018 12:16 am
ovarwa wrote:
Thu Nov 15, 2018 7:36 pm
What are your issues with fleet management?
Lots, but just to give a couple examples of its... unwieldiness:

I have a fleet of 100 ships on system A (let's assume they're the same type), and to react to a double threat I want to send the 70 less damaged ones to system B and the rest to system C. In a game like Stars!, I could do this in a handful of clicks. How many does it take in FO?
Too many!
I do find that clicks and hand movements add up, so you might have noticed some I take for granted. (I think *you* are taking the stuff you consider 'nothing' for granted.
Hardly. What I'm saying is, while there are certainly places where you could divide the number of required mouse clicks by a factor of, say, about two (and that would indeed be an improvement), in the examples above I'm talking about a factor of 100 or more. It doesn't even compare.
Except that I almost never need to do your admittedly bad scenario, whereas I have to go through sitrep every turn.

Anyway,

Ken

defaultuser
Vacuum Dragon
Posts: 630
Joined: Wed Aug 26, 2015 6:15 pm

Re: Finally tried 0.4.8...

#65 Post by defaultuser » Fri Nov 16, 2018 7:32 am

ovarwa wrote:
Thu Nov 15, 2018 7:36 pm

Hmm. For fleet management, the features that would save me the most clicks involve grouping all ships of a type together and pulling out all injured ships in a fleet into a separate fleet.
You can already do that.

For the first, use context->Merge System Ships Into Fleet, followed by context->Split Ships Into Fleets For Each Design.
For the second, use context->Split Damaged Ships From Fleet
For production, I don't like the way I never know where the production window will open; I often have to scroll to what I want. A "most likely" pane in that window would be helpful, containing, say, the last 2 ships I've build there and the 3 newest applicable buildings. (If the place in an outpost, it should include the best colony I can build there.)
It will scroll to the position closest to the position in last previous production window you had open. The problem is that the lists often don't match You get the most scroll problems switching from planet to planet and doing different things. It doesn't try to predict what you should do.

ovarwa
Space Kraken
Posts: 106
Joined: Thu Dec 11, 2014 3:22 am

Re: Finally tried 0.4.8...

#66 Post by ovarwa » Mon Nov 19, 2018 7:33 pm

Hi,
defaultuser wrote:
Fri Nov 16, 2018 7:32 am
ovarwa wrote:
Thu Nov 15, 2018 7:36 pm

Hmm. For fleet management, the features that would save me the most clicks involve grouping all ships of a type together and pulling out all injured ships in a fleet into a separate fleet.
You can already do that.

For the first, use context->Merge System Ships Into Fleet, followed by context->Split Ships Into Fleets For Each Design.
For the second, use context->Split Damaged Ships From Fleet
So I can; thanks!

Anyway,

Ken

defaultuser
Vacuum Dragon
Posts: 630
Joined: Wed Aug 26, 2015 6:15 pm

Re: Finally tried 0.4.8...

#67 Post by defaultuser » Tue Nov 20, 2018 6:19 am

Be aware that when damaged ships are split off, the resulting fleet is "hidden". I still don't like that and would like it to be a start-up option.

User avatar
Geoff the Medio
Programming, Design, Admin
Posts: 12453
Joined: Wed Oct 08, 2003 1:33 am
Location: Munich

Re: Finally tried 0.4.8...

#68 Post by Geoff the Medio » Tue Nov 20, 2018 8:46 am

By "hidden", do you mean passive fleet mode, in which they don't try to start combat and try to stay stealthy?

defaultuser
Vacuum Dragon
Posts: 630
Joined: Wed Aug 26, 2015 6:15 pm

Re: Finally tried 0.4.8...

#69 Post by defaultuser » Tue Nov 20, 2018 5:25 pm

Geoff the Medio wrote:
Tue Nov 20, 2018 8:46 am
By "hidden", do you mean passive fleet mode, in which they don't try to start combat and try to stay stealthy?
Yes, the mode that says: "Hide. Fleet will attempt to stay hidden; it will not blockade or initiate combat."

ovarwa
Space Kraken
Posts: 106
Joined: Thu Dec 11, 2014 3:22 am

Re: Finally tried 0.4.8...

#70 Post by ovarwa » Tue Nov 20, 2018 10:50 pm

defaultuser wrote:
Tue Nov 20, 2018 6:19 am
Be aware that when damaged ships are split off, the resulting fleet is "hidden". I still don't like that and would like it to be a start-up option.
I noticed that, and do not like it either.

It would be useful to be able to select a ship in a fleet and split it off into a new fleet along with all other ships at least as damaged. (Or at healthy.)

For example, I might not really want to put my ship that lost only 1hp with the ships that only have 1hp remaining.

Anyway,

Ken

ovarwa
Space Kraken
Posts: 106
Joined: Thu Dec 11, 2014 3:22 am

Re: Finally tried 0.4.8...

#71 Post by ovarwa » Wed Nov 21, 2018 10:16 pm

Hi,

AI: A new low! A system guarded by a Maintenance Ship has been within sensor range of the Cray for perhaps 75 turns. They have a few warships on a world 1 turn away, and can also send ships from outside my own sensor range. Periodically, they send increasingly silly task forces to camp out on and get destroyed by the MS, such as an unshielded warship, the same guarded by a scout, the same guarded by an outpost ship, until at last, in a sweeping gesture of epic grandeur, an outpost ship guarded by a colony ship. I feel dirty just seeing this.

Speaking of AI: No retreat, no surrender! They send a reasonable (for once) force to a world of mine, but I see it coming and it isn't nearly as reasonable by the time it arrives. Not so bad. The combat ends with all of my ships lightly damaged, and all of their ships destroyed, except for two heavily damaged ships. They do not have another wave of incoming attackers to perhaps support in glorious death. They *could* retreat from whence they came for repairs, but no, they stick around to be destroyed on round one for no benefit.

Anyway,

Ken

ovarwa
Space Kraken
Posts: 106
Joined: Thu Dec 11, 2014 3:22 am

Re: Finally tried 0.4.8...

#72 Post by ovarwa » Thu Nov 29, 2018 8:11 pm

Hi,

A surprise: I only just noticed that non-ultimate engines fit into core *and* internal slots, not just internal slots. This further solidifies the SG hull as the best. It has enough structure to take a hit or two from a big planetary weapon, it has enough slots, and that core slot can be used to overcome speed, its only point of meh. What more do you really need?

Easy AI: It would be really nice if the AI didn't send colony and outpost ships to system that it *knows* are defended. It would be nice if AI moved unarmed ships out of harm's way when it *knows* harm is incoming. It would be nice if the AI did something sane when it *knows* that it has a greatly superior fleet to an oppenent's within 1 turn of travel... or the opposite. (By *knows*, I mean that the situation is within scanner distance and can be demonstrated rather than guessed. Some guesses would also be nice, such as remembering that last turn an opponent had a big fleet within scanner range, and the military strength of that opponent has not gone down, so there must still be a big enemy fleet 1 turn out of scanner range.... but that's a step past easy.)

Anyway,

Ken

User avatar
Dilvish
AI Lead, Programmer
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: Finally tried 0.4.8...

#73 Post by Dilvish » Thu Nov 29, 2018 8:40 pm

ovarwa wrote:
Thu Nov 29, 2018 8:11 pm
Easy AI: It would be really nice if the AI didn't send colony and outpost ships to system that it *knows* are defended.
What version are you playing? It's been quite a while since the AI should have stopped doing that with abandon. A remaining point of weakness is that if it has sent enough warships that it thinks should be enough to clear out the defenses, it could let a colony/outpost ship go ahead and start that last leg, and if it turns out it misjudged the defenses then of course that will often work out badly for it. In general the AI should be safer than it currently is in that regards. But if you see it just sending an un-escorted colony/outpost ship into a known defended location then that shouldn't be happening with current master, please make an Issue with a screenshot and a savegame (preferably from 2-3 turns before the AI sends the ship into doom, so that the savegame could be used to test a fix not just verify the problem).
It would be nice if AI moved unarmed ships out of harm's way when it *knows* harm is incoming.
Morlic recently posted an experimental patch for that, though my recollection is that there were still some bugs with it. Anyways, I think you can expect some improvement on that front to make it into master in the relatively near future.
It would be nice if the AI did something sane when it *knows* that it has a greatly superior fleet to an oppenent's within 1 turn of travel... or the opposite. (By *knows*, I mean that the situation is within scanner distance and can be demonstrated rather than guessed.
Perhaps I am just getting mixed up with planning, but my recollection is that It does have some code that lets it make moves like that at least sometimes, but I agree that clearly a fair bit more improvement is needed. It has a strong tendency to get too hung up on defending a position by sitting in place.
Some guesses would also be nice, such as remembering that last turn an opponent had a big fleet within scanner range, and the military strength of that opponent has not gone down, so there must still be a big enemy fleet 1 turn out of scanner range.... but that's a step past easy.)
That is actually something that has been implemented for a very long time. It might be that whatever you observed really stemmed from the AI not properly coordinating the arrival of multiple fleets. If you think you see it forgetting that a big fleet was present just a turn or two ago, then please, as above, make an issue with screenshot and again with a savegame preferably from a couple turns or so before the big mistake.
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

ovarwa
Space Kraken
Posts: 106
Joined: Thu Dec 11, 2014 3:22 am

Re: Finally tried 0.4.8...

#74 Post by ovarwa » Fri Nov 30, 2018 4:34 am

Hi,
Dilvish wrote:
Thu Nov 29, 2018 8:40 pm
ovarwa wrote:
Thu Nov 29, 2018 8:11 pm
Easy AI: It would be really nice if the AI didn't send colony and outpost ships to system that it *knows* are defended.
What version are you playing?
Whatever the released stable 0.48 is.
It's been quite a while since the AI should have stopped doing that with abandon.
I don't know how to quantify 'abandon', but it happens. Note my earlier post from last week along the same lines.
A remaining point of weakness is that if it has sent enough warships that it thinks should be enough to clear out the defenses, it could let a colony/outpost ship go ahead and start that last leg, and if it turns out it misjudged the defenses
It should *never* misjudge the defenses of a Sentry, Sentinel or Maintenance Ship. So if that's what is happening, something fundamental is wrong. Because judging defenses doesn't get simpler than that. If there are any defenses, of course!
then of course that will often work out badly for it. In general the AI should be safer than it currently is in that regards. But if you see it just sending an un-escorted colony/outpost ship into a known defended location then that shouldn't be happening with current master, please make an Issue with a screenshot and a savegame (preferably from 2-3 turns before the AI sends the ship into doom, so that the savegame could be used to test a fix not just verify the problem).
Not unescorted! No, no. Escorted by a *scout*, yes. Escorted by a ship that will take 10 or more turns to kill a Maintenance Ship, yes. Or by a ship that will get killed *by* the maintenance ship over 5 or more turns, yes.

But not quite unescorted.
Some guesses would also be nice, such as remembering that last turn an opponent had a big fleet within scanner range, and the military strength of that opponent has not gone down, so there must still be a big enemy fleet 1 turn out of scanner range.... but that's a step past easy.)
That is actually something that has been implemented for a very long time. It might be that whatever you observed really stemmed from the AI not properly coordinating the arrival of multiple fleets. If you think you see it forgetting that a big fleet was present just a turn or two ago, then please, as above, make an issue with screenshot and again with a savegame preferably from a couple turns or so before the big mistake.
Hmm. It's hard to know whether it is forgetting about my big fleet, or whether it has decided that something else is more important, or whether it doesn't have a good understanding of how to gauge the strength of a force. Although I'm quite sure the last is true: A lot of the AI behavior is explained by its failure to assess the strength of forces, leading it to attack when it shouldn't, not attack when it should, etc... although not all.

As for the AI coordinating fleets to arrive at the same time, that happens almost never, to the point where I am utterly stunned when AI fleets from different places converge on the same location. Often, AI fleets *from the same location* arrive in dribs and drabs. So if there is AI code that coordinates this, it either isn't being called or is buggy.

Anyway,

Ken

User avatar
Dilvish
AI Lead, Programmer
Posts: 4768
Joined: Sat Sep 22, 2012 6:25 pm

Re: Finally tried 0.4.8...

#75 Post by Dilvish » Wed Dec 12, 2018 7:54 pm

ovarwa wrote:
Fri Nov 30, 2018 4:34 am
Whatever the released stable 0.48 is.
For folks who are active on the boards here we encourage playing the most recent test versions (threads with download links are at the top of the General Discussion Board). This week's test builds include a bug fix related to that out-of-sight-enemy AI issue that was bothering you, hopefully you'd find the handling much better now.
If I provided any code, scripts or other content here, it's released under GPL 2.0 and CC-BY-SA 3.0

Post Reply