Report - Current Bugs and Issues

[Watsong]

Autocombat with an armed Scout ship versus a defenseless outpost results in defeat for the Scout(s), as they auto-retreat from combat...

[zolobolo]

Error pops up every turn during auto battle between two AI controlled fleets

The error seems to be due to a destroyed Light Cruiser of mine which one of the AIs has captured some time ago and which was apperantly captured by the other AI soon after that and now seems to be in the limbo of existence and non-existence

Game uploaded as: game_7363

The game I have uploaded has been switched to the inker faction which currently owns the Light Cruiser in question and produced a crash when I try to play the battle manually

[harpy eagle]

Autocombat with an armed Scout ship versus a defenseless outpost results in defeat for the Scout(s), as they auto-retreat from combat...

That sounds like something that should not happen, but on the topic of retreating from combat, I'm becoming more convinced that the approach I took for the AI deciding when to retreat from combat is flawed. There's no static threshold of force ratio that really works in all situations.

I have a feeling that mixing force ratio together with losses might work better. e.g. if our fleet is outpowered by 3:1 ratio, and we see that we are taking losses accordingly, then we trust the fleet power ratio more as a metric and use it to decide to retreat. But if we see we are inflicting even losses or perhaps disproportionate losses, then we trust the fleet power ratio less.

Of course to do that we would have to be able to store a history of the fleet power in a way that would persist across turns...

[zolobolo]

Autocombat with an armed Scout ship versus a defenseless outpost results in defeat for the Scout(s), as they auto-retreat from combat...

That sounds like something that should not happen, but on the topic of retreating from combat, I'm becoming more convinced that the approach I took for the AI deciding when to retreat from combat is flawed. There's no static threshold of force ratio that really works in all situations.

I have a feeling that mixing force ratio together with losses might work better. e.g. if our fleet is outpowered by 3:1 ratio, and we see that we are taking losses accordingly, then we trust the fleet power ratio more as a metric and use it to decide to retreat. But if we see we are inflicting even losses or perhaps disproportionate losses, then we trust the fleet power ratio less.

Of course to do that we would have to be able to store a history of the fleet power in a way that would persist across turns...

I think the logic you have applied makes sense but it does have some extreme cases
I havew also run into siutations where I send a single heavy cruiser with railguns to seige a planets defenses (since the AI rarely uses rockets or heavy guns) but find my cruiser retreating due to being "overpowered" altogugh it could easily siege the planet from a safe distance

What might thus help is consideration of range: the potential damage output could be weighted against its range. A fleet with huge number of PD guns should be expected to be overpowered by a few heavy gun heavy cruisers as they simply start firing sooner and the same goes for missiles and torps - this would solve the siege retreat issue but not sure abouth the scout: since the defensless outpost does not have guns does your calculation also take into account hit points?

If that it the case, then it should help there as well as the "power" of the unarmed station would be zeroed out due to its 0 range

[sven]

... on the topic of retreating from combat, I'm becoming more convinced that the approach I took for the AI deciding when to retreat from combat is flawed. There's no static threshold of force ratio that really works in all situations.

I'm not convinced it's fundamentally flawed I don't think the retreat heuristic needs to be perfect, it just needs to be good enough that the AI appears to be making choices that are something like rational. Running from a fight that you could win due to balance of weapon types isn't obviously irrational behavior -- it's just a case where the AI admiral is showing an imperfect understanding of tactics.

I think the biggest problem I've noticed with retreating, atm, is that if your fleet does sufficiently outmatch the combat power of the enemy fleet, your opponents will just never stand and fight. That means that if you have the ships necessary to win a war, you'll sometimes chase the enemy fleet through their whole empire, only to have them retreat from the last battle for their last planet, and then disappear from the map. In some sense, this is entirely rational behavior -- those final battles may be honestly hopeless for the AI. But it's anticlimactic. As a player, I'd like to see the AI decide that at some point, it needs to stand and fight, regardless of the odds.

This is probably not that hard to make happen... CombatAI_Method:general_retreat_threshold could probably be tweaked to return a threshold that takes into account the strategic situation, rather than just maxing out at 5 in the case of a planet defense.

[Chasm]

I have the opposite occur. If there are enemy military forces defending a system, they will NEVER retreat, even if they are defending just an outpost. This applies only to combat hulls, not unarmed ships.

[zolobolo]

I have not yet noticed issues with the AI not putting up a fight: since it requires an overwhelming force for them to retreat from defending their planet, and since I am mostly hitting them where there is no defense - but I can see the potential issue there

One solution would be to have the AI consdier its planets when defending: if top 10% most valuable planet: defend agains higher force multiplier then usual, and for the most valuable planet defend at all costs except if civilian ship - these should always retreat). This seems like a minor tweak in the function adding an additional varable to it: the value of the planet being defended

But the issue might also be addressed via the strategic AI script: frankly I think that a strategy solely involving defense is doomed from the start and it is much motr important to:
1. Sue for peace as soon as it is obvious that you are going to be on the short end of a war. Meeting fleets outclassing your entire aramda and loosing valuable planets is already beyond this point
2. Best defense is attack: a considerable portion of the AI fleet should be busy attacking as raiding undefended or lightly defended planets is hugely effective in here as well as blockading and thus reducing the population/moral + preventing buyout witohuth any losses. Harasing the enemy is a great way of detracting enemy forces from attacking your territory and eases the pressure from defending forces. e.g.: the cheapest of destroyes can pose a consdierable threat when bombing and blockading a whole system by itself, while when serving as a defender, its value is negligable. Only downside of this behaviour is that it would be annoying for the player if they do not play attention to defending all systems in enemy range

If the above two are given, I don't see any issues with the current retreat behaviour as even if facing a much larger force which is invading your planets it still makes sense to concentrate on the above

[Dragar]

I usually find that the AI retreats pretty rationally - I also retreat when my fleet is clearly outmatched! While that might lead to some level of frustration for the player, the solution is just to divide up your fleet and strike in multiple locations. This strategy is very effective, as others said above.

Not only is it effective, it makes combat much more small-scale and interesting, and improves balance of PD vs missiles (the latter get out of hand in huge battles, as PD is local while missiles global). It also makes local defenses much more important.

The AI does not really raid in the same way. It would be nice if it did - and make stationary defenses much more important.

When down to a critical world - the last big planet, or homeworld, for instance - the AI might want to fight to the death.

[Mithramuse]

Having a good run through a game, racking up some technologies as I fight off marauders... completed Bombers technology and got 14% towards a "Hidden Bases" technology. Ooh, what's that? I ask.

Descriptive text of "Hidden Bases" states, "This tech should currently be hidden from everyone, and non-discoverable."

Hmm.

Figured I should post that here, just in case. Didn't see any other comments about it when I did a forum search... might just be that I've let this marauder base along for a long time as an AI made an outpost in it, and I had other fish to fry elsewhere while driving off the occasional battleship attack on one of my outposts.

[nathanebht]

Description for Magazine Bays was "Missing info for...". Partial salvage from battling the Yoral.

[zolobolo]

When transporting a single pop unit from one planet to another as the Phidi a ship with two cargo holds is selected instead of a transport ship with a single hold (though such ships are also available for assignment)

Though this scenario should only affect Phidi and only when a single pop is moved, lower capacity transports should be prioratized when the ship is selected else there is a needless income drop from the movement (one cargo hold capacity is needlessly removed from hte trade pool)

[zolobolo]

Noticed that Marauders do not offer slave types that can be found in their direct vicinity. Is this still due to the Marauders not attacking the AI?

If yes, this might slowly be lifted now that the Marauders have been nerfed a bit

[nweismuller]

game_7526: Once again running into the old bug where Marauders get slaves from species they've never been hostile to. Phidi slaves present on Deneb II; I immediately paid off the Marauders on contact and have never had my populations captured by enemies in war (since I've never been at war).

[AMX]

Noticed that Marauders do not offer slave types that can be found in their direct vicinity. Is this still due to the Marauders not attacking the AI?

If yes, this might slowly be lifted now that the Marauders have been nerfed a bit

game_7526: Once again running into the old bug where Marauders get slaves from species they've never been hostile to. Phidi slaves present on Deneb II; I immediately paid off the Marauders on contact and have never had my populations captured by enemies in war (since I've never been at war).

I find the juxtaposition of these two posts somewhat ironic
Are Marauders supposed to have slaves based on their neighbours, or not?

[sven]

I find the juxtaposition of these two posts somewhat ironic
Are Marauders supposed to have slaves based on their neighbours, or not?

I changed the rules yesterday, so both posts are actually accurate, they just refer to different versions of the game. Going to continue experimenting with these rules, as I still think there are some little improvements I can make here.

Lore wise, I don't think it's wholly wrong for Marauders to end up with slaves from neighboring empires that they're at peace with (there's probably enough of a criminal underworld in our universe to justify that happening occasionally). But, I also think that when it's the /player's/ race that shows up as slave population, that can seem particularly wrong. The compromise that I suspect works best is to say that Marauders can spawn slave populations from neighboring empires, but that the player's empire gets a special exemption. I'll probably patch that rule change in later today.