Post reply

Warning: this topic has not been posted in for at least 120 days.
Unless you're sure you want to reply, please consider starting a new topic.

Note: this post will not display until it's been approved by a moderator.

Name:
Email:
Subject:
Message icon:

shortcuts: hit alt+s to submit/post or alt+p to preview

Please read the rules before you post!


Topic Summary

Posted by: Steve Walmsley
« on: November 28, 2021, 12:15:33 PM »

I think this is a fair summary. I am currently coding box-launcher FAC flotillas for NPRs, so it won't be a major step from there to carriers when I get around to it. I might also look at reduced-size launchers for major warships. The scouting and anti-scouting role is trickier, although some form of random patrolling or buoy deployment might work.

Thank you for your work!

I can add information from my observations:
1. The most long-term is nevertheless for the AI ​​to unite its fleet before the main battle.
2. I noticed one more drawback:
Above the enemy planet was a missile defense destroyer, with a very powerful anti-missile defense. And on the planet STO who, when trying to shoot down these ships with missiles, defended them. In total, they were able to shoot down 300 of my good fast rockets at a speed of 37000 km / s.
I am definitely sure that both destroyers and STOs fired at the missiles.

But after I destroyed these destroyers with beam ships, when attacked with STO missiles, they did not shoot them down at all. What conclusion can I draw ...? Perhaps the problem is either with the orders, or (which is quite likely) - with the absence of an active sensor with a resolution of 1.

These are both already in the next version. Its probably worth reading through the change log for v2.0.
http://aurora2.pentarch.org/index.php?topic=12523.0
Posted by: Entaro
« on: November 28, 2021, 11:41:49 AM »

I think this is a fair summary. I am currently coding box-launcher FAC flotillas for NPRs, so it won't be a major step from there to carriers when I get around to it. I might also look at reduced-size launchers for major warships. The scouting and anti-scouting role is trickier, although some form of random patrolling or buoy deployment might work.

Thank you for your work!

I can add information from my observations:
1. The most long-term is nevertheless for the AI ​​to unite its fleet before the main battle.
2. I noticed one more drawback:
Above the enemy planet was a missile defense destroyer, with a very powerful anti-missile defense. And on the planet STO who, when trying to shoot down these ships with missiles, defended them. In total, they were able to shoot down 300 of my good fast rockets at a speed of 37000 km / s.
I am definitely sure that both destroyers and STOs fired at the missiles.

But after I destroyed these destroyers with beam ships, when attacked with STO missiles, they did not shoot them down at all. What conclusion can I draw ...? Perhaps the problem is either with the orders, or (which is quite likely) - with the absence of an active sensor with a resolution of 1.
Posted by: nuclearslurpee
« on: November 28, 2021, 10:29:51 AM »

I think this is a fair summary. I am currently coding box-launcher FAC flotillas for NPRs, so it won't be a major step from there to carriers when I get around to it. I might also look at reduced-size launchers for major warships. The scouting and anti-scouting role is trickier, although some form of random patrolling or buoy deployment might work.

I would love to see NPRs use missile sensor buoys at jump points to monitor traffic and track enemy fleet movements toward core systems. This would hopefully help them a lot with concentrating a fleet to deliver battle to an invader.

I've just updated NPR design philosophy to include the option of reduced-size launchers.

This and the preceding are very welcome changes!  ;D
Posted by: Steve Walmsley
« on: November 28, 2021, 08:17:21 AM »

I've just updated NPR design philosophy to include the option of reduced-size launchers.
Posted by: Steve Walmsley
« on: November 28, 2021, 06:13:14 AM »

Aurora is an extremely complex game and the AI actually are pretty decent if you use a few constriction in your games. But there are a few flaws especially with their design templates. Outside of the ECCM bug the NPR never deploy carriers and it is not that good at scouting or defend against human scouting. The AI also always deploy full size missile launchers which is rather inefficient and relatively easy to defend against with regular beam PD and a small number of AMM. The NPR practically is not able to defend against fighter strikes very well either due to lack of scouting and weapons to defend against them.

As long as you play into the NPR strategies the NPR are decent obstacles, especially if you start at pre-TN and use slow tech progression.

I think this is a fair summary. I am currently coding box-launcher FAC flotillas for NPRs, so it won't be a major step from there to carriers when I get around to it. I might also look at reduced-size launchers for major warships. The scouting and anti-scouting role is trickier, although some form of random patrolling or buoy deployment might work.
Posted by: Entaro
« on: November 27, 2021, 08:33:09 PM »

Aurora is an extremely complex game and the AI actually are pretty decent if you use a few constriction in your games. But there are a few flaws especially with their design templates. Outside of the ECCM bug the NPR never deploy carriers and it is not that good at scouting or defend against human scouting. The AI also always deploy full size missile launchers which is rather inefficient and relatively easy to defend against with regular beam PD and a small number of AMM. The NPR practically is not able to defend against fighter strikes very well either due to lack of scouting and weapons to defend against them.

As long as you play into the NPR strategies the NPR are decent obstacles, especially if you start at pre-TN and use slow tech progression.
HM. I also don’t use aircraft carriers, I don’t practically use fighters, I deploy full-fledged rocket launchers :) I guess this is fair to AI. To be on the safe side, I increased the difficulty of the game to 120%. Suddenly it will help.

For the first time, by the way, I was faced with a situation that my missiles cannot solve.
Enemy mother planet with 4 missile defense destroyers in orbit. In fact, they were able to easily shoot down all of my 200 missiles with their beam weapons. Marvelous.
I'll try to bring the other half of the rocket fleet and fire a salvo of 400 missiles, but I'm not sure what will help :) If it doesn't help, I'll have to use raiders with plasma carronades, but the last time I flew them to this planet, they fired at me quite powerfully.
Posted by: nuclearslurpee
« on: November 27, 2021, 07:36:13 PM »

The AI also always deploy full size missile launchers which is rather inefficient and relatively easy to defend against with regular beam PD and a small number of AMM.

This is to me honestly one of the biggest tactical flaws of the NPRs, along with inability to handle basically any missile attack that isn't also from full-size launchers (box launchers are the worst, but even something like 33% size launchers will run over the NPR fleets). Given what a core part of gameplay missiles are, and how extensive and flexible the missile-related mechanics and systems are, it is a big shame IMO that only the most basic missile tactics are "fair" against the NPRs because of poorly they use and defend against missiles. If you play primarily with beam-armed ships then things are generally more fair, but most of the fun missile tactics have to be reserved for use against other player races.
Posted by: Jorgen_CAB
« on: November 27, 2021, 07:22:41 PM »

Aurora is an extremely complex game and the AI actually are pretty decent if you use a few constriction in your games. But there are a few flaws especially with their design templates. Outside of the ECCM bug the NPR never deploy carriers and it is not that good at scouting or defend against human scouting. The AI also always deploy full size missile launchers which is rather inefficient and relatively easy to defend against with regular beam PD and a small number of AMM. The NPR practically is not able to defend against fighter strikes very well either due to lack of scouting and weapons to defend against them.

As long as you play into the NPR strategies the NPR are decent obstacles, especially if you start at pre-TN and use slow tech progression.
Posted by: Entaro
« on: November 27, 2021, 04:58:04 PM »

HM..
Do I understand correctly that in order to play honestly against AI, you must either not use ECM yourself, or write the above line in certain files?
Posted by: Jorgen_CAB
« on: November 27, 2021, 04:35:55 PM »

As for ECCM I also hope NPR AMMs also start using ECCM as well. Missiles with decent ECM are more or less impossible by the NPRs to intercept with their AMM. Even at around ECM 3 the AI AMM become almost useless.
Posted by: Kristover
« on: November 27, 2021, 03:56:00 PM »

much words

It is worth noting that at least a partial fix may come in v2.0, as per Steve post here the AI will try to combine its fleets in a system instead of sending piecemeal forces, so offensive ships should be accompanied by anti-missile escorts assuming these are present in the system. Frankly if this change works its impact in making NPRs reasonable opponents cannot be overstated I think although how well it works will remain to be seen.

I can think of at least two of my games that would have ended early if the AI had combined their fleet before coming at me.
Posted by: Droll
« on: November 27, 2021, 03:54:06 PM »

much words

It is worth noting that at least a partial fix may come in v2.0, as per Steve post here the AI will try to combine its fleets in a system instead of sending piecemeal forces, so offensive ships should be accompanied by anti-missile escorts assuming these are present in the system. Frankly if this change works its impact in making NPRs reasonable opponents cannot be overstated I think although how well it works will remain to be seen.

I know it's been discussed before but for me where the NPRs really falloff is in the empire building aspect and not the tactical side, I think the ECCM bug is major but that's the only massive thing that I've noticed tactically.

I don't ever expect them to match my meager tactical genius but if the AI, esp. the generated NPRs get a bit better at building empires then I am gucci.
Posted by: nuclearslurpee
« on: November 27, 2021, 02:18:20 PM »

much words

It is worth noting that at least a partial fix may come in v2.0, as per Steve post here the AI will try to combine its fleets in a system instead of sending piecemeal forces, so offensive ships should be accompanied by anti-missile escorts assuming these are present in the system. Frankly if this change works its impact in making NPRs reasonable opponents cannot be overstated I think although how well it works will remain to be seen.
Posted by: Entaro
« on: November 27, 2021, 01:56:47 PM »

The AI designs and builds ships with newer technology as it progresses. There are other options for missile defence besides railguns and not every AI fleet will have effective AM protection because some have different roles.

The AI looks at every shot you fire and everything your ships do and adjusts accordingly, using same tactical intelligence information that is available to the player. It is a LOT more sophisticated than your suggestion above. However, as with any AI, it is never going to come close to a human player in a tactical situation because a human can weight different factors in a given situation and plan accordingly. Also, the more complex the game, the more difficult to create an AI that can adjust to any situation. If I made Aurora much simpler, the AI would handle it far better.

Consider how basic the rules are for Chess or Go or Poker and consider how much effort and how many years were required to build an effective computer opponent. I suggest playing the game longer until you have encountered a lot more situations and then think about the AI code that would be required to handle all of them given the huge variety of different factors and potential systems involved.
I understand. Sorry, I didn't want my error message to sound like a rebuke to you personally! I understand perfectly well that Aurora is a very difficult game, and it is almost impossible to make an adequate AI in it that would be comparable to a human in abilities.

Off-Topic: show
I previously played Gary Grigsby's War in the East - a game about the Second World War, where each support division / regiment is a separate unit with many parameters, and a map of the Second World War (Eastern Front) where every 10km (+ -) is a separate hex with its own parameters. It is also a very complex game, and there it was also almost impossible to create an AI even approximately comparable to a human.


Simply, specifically in this situation, it seems to me that we are talking about some kind of error in logic. I'm sure it doesn't have to be that way. I understand that AI is not perfect, that it can be stupid, separate its fleets, not calculate forces, design ships not in the most optimal way ...
But when the AI, after the conclusion of a truce, does not react to the fact that I destroy its ships with missiles, or when it sends fleets that have no anti-missile defense at all (despite the fact that it has separate missile defense ships!) - to my ships with missiles - this is very strange.

The point is that the AI ​​has designed a certain type of ship:
having: 17 Railgun DMG 3 * 4 80904 km ROF 150.
And almost all of its combat fleets consist only of ships of this type.
I don't know what kind of equipment they have inside, but they are definitely not capable of firing at my missiles. Even on slow (16000 km / s).

Either there is a mistake that the AI ​​does not install anti-missile sensors on these ships, and therefore is simply not able to see them, or the AI ​​does everything correctly in terms of ship templates (having separate missile defense ships - I also do this, and do not put sensors 1 on my main warships), but does not add ships of the main combat class - missile defense destroyers to its fleets.
Posted by: Droll
« on: November 27, 2021, 01:45:25 PM »

I also found a bug in NPR automated ship design that prevented them from using ECCM. That is fixed for v2.0.

This explains so much why ECM dominates massively. I think this fix alone is going to have a large effect on generated NPR difficulty moving forward.

If you want to do a hot-fix, go into DIM_AutomatedClassDesign and ensure there is a 1 in the ECCM field for every design that has a 1 in the ECM field, except for the military tanker and scout.

Thanks, this is the query I used to help find everything (could probably make an update query instead):

Code: [Select]
select DesignID, ECCM
from DIM_AutomatedClassDesign
where (ECM = 1 and
      (not DesignID = "Scout" and
       not DesignID = "Military Tanker"));

Question regarding this hotfix for Steve but it has to be spoilered:
The invader scout has ECCM set to 0, since it is a type of scout I left it at 0 but do the invaders have special armed scouts that should have ECCM?