Hard Light Productions Forums
General FreeSpace => FreeSpace Discussion => Topic started by: General Battuta on May 09, 2011, 12:41:53 pm
-
Looking at the game universe, and taking AI Class as a representative of pilot or crew skill, excluding multiplayer (we will take this as a universe without players, just the FreeSpace AI) but accepting empirical data from tests, do you consider pilot skill or the capabilities of a spacecraft to be more important to a unit's performance?
**** polls, we don't need a poll, let's just discuss it.
As a bonus include Fury AI and its capabilities.
-
are we excluding the part of pilot skill that directly affects fighter capabilities or are we assuming this is a sensible ai.tbl which does not alter them
-
are we excluding the part of pilot skill that directly affects fighter capabilities or are we assuming this is a sensible ai.tbl which does not alter them
We will exclude BALLS OF STEELE and define the maximally skilled pilot as one who can operate his craft and weapons at the limits of their mechanical performance (i.e. as defined in ships.tbl). We will also take as a given that most pilots are unable to achieve this level of performance since the default AI classes are well below this; however, we will ignore stupid retail limitations like max attackers on one target.
-
Pilot skill and partially, the fighter's stats. If you have a state-of-the-art fighter with super-hyper-ultra fast speed, if the AI using it is Coward, it'll be pretty much useless, maybe for running away. If you put a decent AI in an average space superiority fighter, it can very well beat a less intelligent AI if it hasn't already ran away.
For capital ships...well. Balls of Steele + Standard Flak. AI profile dictates aiming, refire rate and target priority for turrets, if I recall correctly. So a good AI can do better with a Fenris in shooting down a bomber wing, than a horrific AI with an Aeolus.
-
This topic would probably be better discussable if we were to do some tests. Like a wing of Herc II's vs a wing of Ares' and alternate the a.i. profiles each round or something along these lines
-
This topic would probably be better discussable if we were to do some tests. Like a wing of Herc II's vs a wing of Ares' and alternate the a.i. profiles each round or something along these lines
I think that was the general (ar ar ar) idea.
-
Simply comparing AI and fighter stats is not a complete picture I think. That's like having to choose between Michael Jordan, David Beckham and Verne Troyer on who the better sportsman is: it all comes down on if they're going to play basketball, football or midget golf.
I think a much larger number of factors is decisive for the effectiveness of a computer controlled fighter in the field. Hitboxes matter a lot: nobody likes to dogfight an SF Dragon. Gun placement is important also probably: I for one never got killed by the primaries of a Seraphim. Also mission placement is extremely decisive: having hostile fighter wings emerging from several angles may proof far more fatal than a heads on approach, unless the heads on approach is an Ares with trebuchets, which all comes down to optimal placement dependant on the role of the hostile craft. If not using the rookiee AI (I find Freespace AI rookies to be unconvincing, everyone's trained at some academy in freespace; yet freespace AI rookies fly like conscripts without training), I'd say:
(http://img715.imageshack.us/img715/1705/graphy.jpg)
-
Ok Sara, although what it would be really interesting was to actually write how a really objective test to these assumptions could be, rather than posting subjective percentages, even if they look very pretty in a cake chart.
-
I'd say:
(http://img715.imageshack.us/img715/1705/graphy.jpg)
Sara's personal opinion is personal. What's your opinion, Luis? Or do you have one? Whilst we wait for test results, we can still opine about our views. :P If we have them.
-
Ok Sara, although what it would be really interesting was to actually write how a really objective test to these assumptions could be, rather than posting subjective percentages, even if they look very pretty in a cake chart.
My point is there is no objective test, neither is choosing better AI over ship stats valid, nor will it matter how skilled the hostile AI is if your faction has a significant tactical advantage such as static defenses. In the end a mission is designed in a specific way and smart placement can make two wings of AI Captains in Myrmidons with the right loadout doing evasive maneuvers far more deadly to an enemy target than two wings of AI Generals Terran Mara's flying banzai together in a straight line (fire two linked Infyrnos at them and they're gone). You can look at which stats, either AI or ship stats make for the dogfighter with the most advantages, but rarely is a mission decided by two groups of fighters having a fair dogfight through eventless contest. There's elements of surprise, sneak attacks, weapons behaviour (EMP, shield drainers, cluster weapons) and even the presence of capital ships.
Luis Dias. Don't bother if you can't motivate or back up your belittling with some kind valid argument or own suggestion, you're just attacking as usual.
-
Don't mind Luis, he's a bit naff.
For fun let's try various ships using various AI settings in a two versus two (or one versus one) fight with Prom S, Subach/Balor, Harpoons and Tempests. Start at 2 kilometers, no terrain.
-
Well, Sara is very precise with 1/20 values to share between factors. I also think that her factors are well chosen. We should view "hitbox size/shape" and "ship stats" as something about ship capabilities, and we only see 30% attributed to AI.
I can't be as precise, and I have the slight suspicion that the exercise itself is empirically too vague to attribute anything.
It depends too much upon the human player's abilities (and capabilities - like the pc they have, the mouse, the joystick, the visuals, etc.), so that a certain share (like Sara's) may not transport itself to another player's experience. If we were to be technical, we would have to pit AI against AI and nuke them out several times to get significant statistics. But that would be missing the point, since AI will not behave like we do, and so they will be reporting different types of abilities than the ones that are actually important to us.
Still, it would be the closest proxy to what Battuta is asking.
If however, we are to make it up with our gut instinct, I'd say they are pretty comparable (AI vs ships). I can take out very easily ships that are fast and capable but with a terrible AI, just as it is easy to take out ships with good AI but terrible specs. But you see the vast number of lurking questions behind this simplistic of a sentence. I'd place my confidence on that sentence in less than the sufficient to warrant significance ;).
-
Ok Sara, although what it would be really interesting was to actually write how a really objective test to these assumptions could be, rather than posting subjective percentages, even if they look very pretty in a cake chart.
My point is there is no objective test. In the end a mission is designed in a specific way and smart placement can make two wings of AI Captains in Myrmidons with the right loadout doing evasive maneuvers far more deadly to an enemy target than two wings of AI Generals Terran Mara's flying banzai together in a straight line (fire two linked Infyrnos at them and they're gone). You can look at which stats, either AI or ship stats make for the dogfighter with the most advantages, but rarely is a mission decided by two groups of fighters having a fair dogfight. There's elements of surprise, sneak attacks, weapons behaviour (EMP, shield drainers, cluster weapons) and even the presence of capital ships.
Luis Dias. Don't bother if you can't motivate or back up your belittling with some valid argument, you're just attacking.
Calm down Sara, I wasn't attacking you at all. I saw what you did and I think that the factors you chose were pretty good. But, as you say, they are subjective, so I don't really see the point of making a pie chart out of it.... with the (good) exception of it being like a personal guidance to your Fredding.
-
Don't mind Luis, he's a bit naff.
For fun let's try various ships using various AI settings in a two versus two (or one versus one) fight with Prom S, Subach/Balor, Harpoons and Tempests. Start at 2 kilometers, no terrain.
So a purely hypothetical and fair group vs group battle?
-
"Remember what the commander said? It's not what you fly, it's how you fly it!" *Derelict
I agree with that. Just look at the beginning of the Great War. Shivans had everything and had to be bathed in Furies to go down, yet they DID go down on several occasions.
-
At least in FS2 Shivan pilots are on average better, IIRC.
-
Yeah, but they have to compensate for shi**** primaries, small secondary bank, and poor weapon placement most of the time
-
Did a few tests. 4 herc II's vs 4 Ares. Starting distance 1700, facing each other. Default loadout for both (so prom r, subach, harpoon and hornet)
Retail captain A.i
Resulted in these scores (ships left):
Herc wing: 0 0 3 3 0 3 0 0
Ares wing: 2 2 0 0 3 0 4 3
Herc II's with BP-Major A.I
Ares' wing with retail captain A.I.
Herc wing: 4 4 3 4 4
Ares wing: 0 0 0 0 0
Quite a clear victor there
The last run I replaced the Ares with Erinyes and gave them a bank of Prometheus S and a bank of Kayser. The Hercs still have the BP-Major A.I. profile (and their default loadout) while the Erinyes retain the retail Captain profile
Herc II wing: 0 0 0 0 0
Erinyes wing: 4 2 2 3 3
Draw your own conclusions
Edit: by battuta's request
Herc II's upgraded with temptest (though they now seem more willing to use hornets instead) and BP-General A.I.
Herc II wing: 0 0 0 3 0 0
Erinyes wing: 1 3 3 0 3 1
Herc II's prom R replaced with prom S
Herc II wing: 0 2 0 0 0 4 0
Erinyes wing: 3 0 2 1 1 0 2
-
SCIENCE
It seems the skill of the pilots is enough to tip the fight in some cases, but the hardware advantage still weighs in favor of the Erinyes.
-
Most of the time because the Herc's get vaporised on the first jousting run.
If they somehow manage to survive that in decent enough numbers they can win the resulting close range dogfight
-
8 Gunpoints lets you win jousting runs.
Thus hardware advantage.
-
So what is really needed is a new ship class, let's call it AREION, with 12 gunpoints and enough power to feed them.
-
So what is really needed is a new ship class, let's call it AREION, with 12 gunpoints and enough power to feed them.
http://www.hard-light.net/wiki/index.php/UEF_Izra%27il
http://www.hard-light.net/wiki/index.php/GTF_Ladiel
ololololol - http://www.hard-light.net/wiki/index.php/VEF_Aoh
-
So what is really needed is a new ship class, let's call it AREION, with 12 gunpoints and enough power to feed them.
http://www.hard-light.net/wiki/index.php/UEF_Izra%27il
http://www.hard-light.net/wiki/index.php/GTF_Ladiel
ololololol - http://www.hard-light.net/wiki/index.php/VEF_Aoh
Lovely. Now we gotta slap Morning Stars on them.
-
@Destiny: And do what? Knock cruisers around? :lol:
-
The only thing that would knock cruisers around would be heavy beam cannons :nervous:!
Okay, I've just made a test. 4 friendly Hercules vs. 4 enemy Erinyes. Unpredictably, the Hercules blasted the crap out of the Erinyes, and the most the enemy could do was either kill one friendly (in the first test) or damage a friendly to the point where he screams "hull integrity failing!" (in the second test).
Is the friendly AI naturally better or something?
Also the Erinyes appeared to be firing only 4 gun banks, I think, though I'm not entirely sure.
-
AI profiles/difficulty level controls if they link their banks.
Play on insane.
-
The only thing that would knock cruisers around would be heavy beam cannons :nervous:!
Okay, I've just made a test. 4 friendly Hercules vs. 4 enemy Erinyes. Unpredictably, the Hercules blasted the crap out of the Erinyes, and the most the enemy could do was either kill one friendly (in the first test) or damage a friendly to the point where he screams "hull integrity failing!" (in the second test).
Is the friendly AI naturally better or something?
Also the Erinyes appeared to be firing only 4 gun banks, I think, though I'm not entirely sure.
Play on insane, use Fury AI to remove stupid IFF differences, make sure the mission has BP2 or equivalent AI profile.
-
Most of the time because the Herc's get vaporised on the first jousting run.
If they somehow manage to survive that in decent enough numbers they can win the resulting close range dogfight
Very fair assessment. Performance wise, the only difference between a Herc 2 and an Ery is that the latter is 10m/s faster.
Can someone do a test on a wing of Herc 2's vs. a cruiser and a wing of Erinyes vs. said cruiser? I've seemingly observed in the past that the Ery just couldn't last long as an assault/strike fighter. This is odd to me; I always felt the Ery was a bigger target, though that's assumably false... It does have a smaller weapons load, but with the guns it shouldn't matter... It has more HP than the Herc 2, though it also has weaker shields. That said, I've got no idea why it seems to be a less survivable fighter wing than a Herc 2 wing in the retail campaign.
-
Most of the time because the Herc's get vaporised on the first jousting run.
If they somehow manage to survive that in decent enough numbers they can win the resulting close range dogfight
Very fair assessment. Performance wise, the only difference between a Herc 2 and an Ery is that the latter is 10m/s faster.
Can someone do a test on a wing of Herc 2's vs. a cruiser and a wing of Erinyes vs. said cruiser? I've seemingly observed in the past that the Ery just couldn't last long as an assault/strike fighter. This is odd to me; I always felt the Ery was a bigger target, though that's assumably false... It does have a smaller weapons load, but with the guns it shouldn't matter... It has more HP than the Herc 2, though it also has weaker shields. That said, I've got no idea why it seems to be a less survivable fighter wing than a Herc 2 wing in the retail campaign.
This is not a ship vs. ship thread, but feel free to open one.
-
To really balance whether or not the difference comes down to AI or Hardware, you'd need to set up something like the following, I'd think:
Herc/Herc2 vs Herc/Herc2 (make sure they are the same in both, matching loadouts), where Friendly has it's AI class to BP and Hostile has Retail.
Captain to Captain and the like or as near equivalent as possible.
(For real regression, then repeat the exercise with the Friendly using the Retail AI and the Hostile using the BP AI just to remove any factors)
You can then vary the ships (but always matching ship to ship and loadout to loadout) however many times you want to gauge the effective levels of AI vs AI when hardware configuration is no longer the concern.
Additionally, if you wanted to adjust AI levels (so that it's no longer a Captain vs Captain of 2 AI behaviours) you could do that as well.
-
AI doesn't perform by its AI class alone but AI profile also plays significant role. Because of this, comparing retail AI vs. custom AI is literally impossible. Why? Because retail AI is not retail AI if you use custom AI profile, the retail AI would receive many of the benefits a custom AI has simply by using custom AI profile. For this reason, any comparison of retail vs. custom AI cannot be done with any decent accuracy because retail AI is no longer retail AI if its behavior is altered by AI profiles.
Now, it IS possible to create ai.tbl and ai_profiles.tbl where all AI behavior alterations are located under specific AI classes. But not quite all AI behavior alterations are supported in ai.tbl, which means the custom AI would not be able to perform at its best.
The bottom line? It is not possible at this time to compare two AI classes that were designed to use different AI profiles. Only AI classes that were designed to use same AI profile are comparable. This also explains the fact that retail AI classes are even able to put up a fight with custom AI classes, either custom AI classes sucked because retail AI profile was used or retail AI classes were better than retail because custom AI profile was used.
-
This also explains the fact that retail AI classes are even able to put up a fight with custom AI classes, either custom AI classes sucked because retail AI profile was used or retail AI classes were better than retail because custom AI profile was used.
retail AI classes custom AI classes, either custom AI classes retail AI profile was used or... retail AI classes were better... retail because custom AI profile was used.
:jaw:
Wow, just imagine if you said this ten times with a motor mouth! :eek2: