Poll: Worst Unit Of The (time interval) Award (IV)

[TechSY730]

Ah ha, non sniper turrets get their firepower divided by 10. (for some firepower evaluations)

That would explain why the AI seems to massively underestimate the effectiveness of turrets.

Yep.  The real question is whether you really want them to be even more skittish about crossing over

I personally do. It seems strange that the AI constantly sends in tiny groups of ships against a large turret ball, and instantly gets killed.

Could that denominator be bumped down to 5 or 4? That should still make them consider that turrets are immobile, but hold off in sending underwhelming fleets against large turret balls.

[Wanderer]

Ah ha, non sniper turrets get their firepower divided by 10. (for some firepower evaluations)

That would explain why the AI seems to massively underestimate the effectiveness of turrets.

Yep.  The real question is whether you really want them to be even more skittish about crossing over

Not really.  They already have the committal problems of a heartbroken romance novelist.

[TechSY730]

Ah ha, non sniper turrets get their firepower divided by 10. (for some firepower evaluations)

That would explain why the AI seems to massively underestimate the effectiveness of turrets.

Yep.  The real question is whether you really want them to be even more skittish about crossing over

Not really.  They already have the committal problems of a heartbroken romance novelist.

So would you like them to not wait for quite so much in general?

My point is that the AI should not be treating turrets as cheaply as they do in comparison to fleet ships.

Now, could the AI wait less long for a build up for everything, including fleet ships, probably. The stalking behavior is a very smart move by the AI, but sadly kind of hurts the flow of the game at it's current magnitude.

I could live with them not underestimating stationary stuff so much, in return for not requiring quite such a large ratio of firepower (like, if they needed 1.25 times the firepower to enter the planet now, make the turret change but only have them require only .8 times the firepower to consider entering)
This is one of those cases where making the AI slightly stupider is offset by the benefit to the game's flow.

[Wanderer]

Ah ha, non sniper turrets get their firepower divided by 10. (for some firepower evaluations)

That would explain why the AI seems to massively underestimate the effectiveness of turrets.

Yep.  The real question is whether you really want them to be even more skittish about crossing over

Not really.  They already have the committal problems of a heartbroken romance novelist.

So would you like them to not wait for quite so much in general?

My point is that the AI should not be treating turrets as cheaply as they do in comparison to fleet ships.

Now, could the AI wait less long for a build up for everything, including fleet ships, probably. The stalking behavior is a very smart move by the AI, but sadly kind of hurts the flow of the game at it's current magnitude.

I could live with them not underestimating stationary stuff so much, in return for not requiring quite such a large ratio of firepower (like, if they needed 1.25 times the firepower to enter the planet now, make the turret change but only have them require only .8 times the firepower to consider entering)
This is one of those cases where making the AI slightly stupider is offset by the benefit to the game's flow.

I don't mind turrets being evaluated properly, but yes, that would require a significant adjustment to their commital values.  A couple of fortresses and a MK II fleet shouldn't threatball 1300 + ships, when 400 will go after turrets.  I agree with that, consistency is horrible.  However, I don't want everything in the universe to just threat-ball, either.   

[Hearteater]

Fortresses are mobile, so I don't think they get 10% firepower, which explains why it is hard to bait the AI into a fortress.

[TechSY730]

Ah ha, non sniper turrets get their firepower divided by 10. (for some firepower evaluations)

That would explain why the AI seems to massively underestimate the effectiveness of turrets.

Yep.  The real question is whether you really want them to be even more skittish about crossing over

Not really.  They already have the committal problems of a heartbroken romance novelist.

So would you like them to not wait for quite so much in general?

My point is that the AI should not be treating turrets as cheaply as they do in comparison to fleet ships.

Now, could the AI wait less long for a build up for everything, including fleet ships, probably. The stalking behavior is a very smart move by the AI, but sadly kind of hurts the flow of the game at it's current magnitude.

I could live with them not underestimating stationary stuff so much, in return for not requiring quite such a large ratio of firepower (like, if they needed 1.25 times the firepower to enter the planet now, make the turret change but only have them require only .8 times the firepower to consider entering)
This is one of those cases where making the AI slightly stupider is offset by the benefit to the game's flow.

I don't mind turrets being evaluated properly, but yes, that would require a significant adjustment to their commital values.  A couple of fortresses and a MK II fleet shouldn't threatball 1300 + ships, when 400 will go after turrets.  I agree with that, consistency is horrible.  However, I don't want everything in the universe to just threat-ball, either.   

Hmm, mantis the request to lower the "firepower ratio" the AI waits for before entering a planet, and relate it to the adjusting how the AI values turret firepower?

[Wanderer]

Ah ha, non sniper turrets get their firepower divided by 10. (for some firepower evaluations)

That would explain why the AI seems to massively underestimate the effectiveness of turrets.

Yep.  The real question is whether you really want them to be even more skittish about crossing over

Not really.  They already have the committal problems of a heartbroken romance novelist.

So would you like them to not wait for quite so much in general?

My point is that the AI should not be treating turrets as cheaply as they do in comparison to fleet ships.

Now, could the AI wait less long for a build up for everything, including fleet ships, probably. The stalking behavior is a very smart move by the AI, but sadly kind of hurts the flow of the game at it's current magnitude.

I could live with them not underestimating stationary stuff so much, in return for not requiring quite such a large ratio of firepower (like, if they needed 1.25 times the firepower to enter the planet now, make the turret change but only have them require only .8 times the firepower to consider entering)
This is one of those cases where making the AI slightly stupider is offset by the benefit to the game's flow.

I don't mind turrets being evaluated properly, but yes, that would require a significant adjustment to their commital values.  A couple of fortresses and a MK II fleet shouldn't threatball 1300 + ships, when 400 will go after turrets.  I agree with that, consistency is horrible.  However, I don't want everything in the universe to just threat-ball, either.   

Hmm, mantis the request to lower the "firepower ratio" the AI waits for before entering a planet, and relate it to the adjusting how the AI values turret firepower?

I'm not going near Mantis until I think over the code and information that Keith posted for a day or two while I debate on resultant problems.  I also want to go into the game and see if I can nail down, or at least get a better feel, for larger threat-ball firepower levels vs. a full turretball.  I want to use an 'everything but the kitchen sink (fortress)' turretball as my baseline for any opinions I might eventually have.

This mechanic isn't something I'd want to adjust friviously or with 'first-glance' opinions.

[Nodor]

I approve of the VAST difference between a "I have 3 fortresses, the AI is never going to enter this planet again" and "I have 200 turrets, yet the AI will come in here and die". 

Effectively, both are set planet defenses and this delta means you don't really want to pair them. 

This enables you to use turrets on a whipping boy/threat bleed front line location, while building a line of fortresses that will restrict the flow of ships deeper into your backfield - minimizing your losses if the wave overwhelms a front line planet.  This also means that I will feel less obligated to move fortresses forward. 

[TechSY730]

I approve of the VAST difference between a "I have 3 fortresses, the AI is never going to enter this planet again" and "I have 200 turrets, yet the AI will come in here and die". 

Effectively, both are set planet defenses and this delta means you don't really want to pair them. 

This enables you to use turrets on a whipping boy/threat bleed front line location, while building a line of fortresses that will restrict the flow of ships deeper into your backfield - minimizing your losses if the wave overwhelms a front line planet.  This also means that I will feel less obligated to move fortresses forward.

Sure its very useful from a gameplay perspective, but one of the goals is trying to make the AI play smart, not necessarily playing in a way that is nice to deal with. (This game is called AI war after all) Having them undervalue turrets is not smart.
Of course, this needs to be tempered with the "fun" goal, where if the AI starts getting too smart, it can start throwing off the pacing and flow of the game. (or at least on lower levels, at 9 and up, I expect Rommel like levels of progress impeding intelligence)

[Nodor]

In my opinion, if the AI needs to be smarter, you need to enable more options.   Advanced Hybrid Hives, Hard Golems and Spirecraft, Astro Trains, all of the Neinzul options, Mining Golems, etc. 

As the amount of stuff you have to deal with multiplies given the compounding effects..  for instance, a CPA, + 3 Exos + 2 waves + mining golems during a hack = BUSY, you get a very nasty management challenge. 

I'm not sure I want much more of an AI brain on top of all that chaos.

[TechSY730]

In my opinion, if the AI needs to be smarter, you need to enable more options.   Advanced Hybrid Hives, Hard Golems and Spirecraft, Astro Trains, all of the Neinzul options, Mining Golems, etc. 

As the amount of stuff you have to deal with multiplies given the compounding effects..  for instance, a CPA, + 3 Exos + 2 waves + mining golems during a hack = BUSY, you get a very nasty management challenge. 

I'm not sure I want much more of an AI brain on top of all that chaos.

I do.

In all seriousness though, even if you stick to base game only stuff, no AI plots, and no hacking, it is my opinion that the AI should still be quite smart (or at least starting at around 6 or 7 difficulty).
I don't really like ideas that intentionally make the AI "dumber" or keep the AI doing dumb stuff just because it is annoying to deal with their behavior when it is smart.
As I said though, some concessions do need to be made to keep the game fun and flowing. It may be that having a very easy, low micro way of baiting the AI to throw away "threat balls" is necessary and even required to keep the game flowing, even if it is necessary to keep the AI "dumb" to do so.
I'm not yet convinced this is the case though.

I am not advocating the turrets just get considered like normal. They still would be considered less powerful than they are to the AI, as the turrets are immobile after all. I am proposing that the magnitude that the AI underestimates turrets is too high right now.
Also, I mentioned that if this change were made, the "firepower ratio" cutoffs for when the AI is willing to enter the planet may need to come down a little, as now the firepower the AI considers the player to have has gone up. (in fact, this may need to happen anyways, as "threat balls" are a bit too prominent currently IMO)

[keith.lamothe]

If the AI were made smart in the "this helps me win" sense, you'd be swimming in MkV vengeance shortly after taking your first planet

[TechSY730]

If the AI were made smart in the "this helps me win" sense, you'd be swimming in MkV vengeance shortly after taking your first planet

One of the reasons why "full blown" AI intelligence in terms of winning is a bad idea. You are right, if the AI truly wanted to win (in the since of eliminating all humans), the smart thing to do would be to free everything on every planet that isn't critical to defend and rush the humans at the very start of the game. But that wouldn't be fun.

However, something as simple as seeing turrets and thinking "Oh turrets, they can hardly hit me, let's rush in!" when in fact those turrets can hit hard is not only dumb, but not a necessary dumb in terms of keeping the game fun and flowing.
Of course, if someone can show me a strong case that having the AI be stupid in regards to turrets compared to other sources of firepower (to a large magnitude) is needed to keep the game fun and flowing, and that all other practical alternative changes to keeping threat balls under control without making the game a grind and keeping the AI smart are not possible or introduce worse problems (aka, give a strong case that the above is a "necessary dumb"), I may reconsider.

[Nodor]

So, my turret defense on my current whipping boy is based on that planet being a former superterminal planet. 

About 75% of my non-sniper turrets on the planet do not have intercept range on any of the wormholes currently leading to the planet with the AI.  About 50% of my turrets do not have intercept on any point the AI is likely to head. (due to the Exo's taking out the Black hole machine they were protecting)

I can move them to make them more effective of course, but the problem with turrets is much the same problem the AI has with guard posts. - immobile objects without overlapping ranges that cannot protect other objects from the edge of ranges can be taken down without impacting the agressive unit.  Blade Spawners, Sentinal Frigates, stealther units, and missile frigates have an effective range advantage such that you would be lucky to inflict 10% casualty damage on those types of units.

Now, good positioning with overlapping fire and protective units (gravity turrets, radar jammer Mk2, tachyons, anti-snipe & anti missile turrets, mines etc.) can make turrets extremely effective.   But I don't think the game should assume the tech purchases in base turret threat evaluation mechanic by the threatball unless those turrets are on the planet the threatball is considering.

[TechSY730]

So, my turret defense on my current whipping boy is based on that planet being a former superterminal planet. 

About 75% of my non-sniper turrets on the planet do not have intercept range on any of the wormholes currently leading to the planet with the AI.  About 50% of my turrets do not have intercept on any point the AI is likely to head. (due to the Exo's taking out the Black hole machine they were protecting)

I can move them to make them more effective of course, but the problem with turrets is much the same problem the AI has with guard posts. - immobile objects without overlapping ranges that cannot protect other objects from the edge of ranges can be taken down without impacting the agressive unit.  Blade Spawners, Sentinal Frigates, stealther units, and missile frigates have an effective range advantage such that you would be lucky to inflict 10% casualty damage on those types of units.

Now, good positioning with overlapping fire and protective units (gravity turrets, radar jammer Mk2, tachyons, anti-snipe & anti missile turrets, mines etc.) can make turrets extremely effective.   But I don't think the game should assume the tech purchases in base turret threat evaluation mechanic by the threatball unless those turrets are on the planet the threatball is considering.

The fact that not all turrets will be able to hit ships as soon as they come in due to their finite range and "stationaryness" should be considered by the AI.
That is why the AI was dividing by 10 in the firepower of turrets, to model this. What I am arguing is that 10 is too high of a denominator, unless people regularly only have 1/10th of their turrets able to hit an attack force at any one time.

What is a good denominator? The problem is that it depends heavily on the turret placement. The more "overlapping" the turret ranges and how many "unavoidable" turrets (that is, turrets that can cover the wormhole the AI will enter from or turrets that are protecting important targets like the command station), the smaller that denominator should be.

Of course, it is impractical for the AI to try to compute "turret coverage" of areas of interest every time they want to enter a planet and need to consider turrets. Thus, the denominator in that function serves as a loose approximation of "turret coverage" of common turret placement configurations. What I am arguing is that the current denominator (10) is a poor approximation of common turret placement patters, even by loose approximation standards.