A "real" AI cheat
- Thread starter The Snug
- Start date
Roland Johansen
Deity
The submarine had a chance to die of 94% if it wouldn't retreat. It's chance to retreat must have been less than that, so it was actually stupid to keep attacking/pillaging while it could very easily lose. It happened to win against the extremely damaged missile cruiser (16 hps at start), but that was just very lucky.
Note that I'm against these repeated retreats, but just because I view the patrol ship as the attacker and thus the pillager shouldn't be allowed to retreat at all.
If a unit with odds of say 80% chance to retreat (which is already extremely high) loses and retreats, then in most cases it's a very bad idea to keep attacking. With such low hitpoints the chance that it will die is much higher than the chance that it will keep retreating until it kills its opponent. Just because the AI keeps doing this and sometimes is lucky doesn't make it a sound tactic. It just means that a stupid tactic is sometimes successful through pure luck.
thadian
Kami of Awakened Dreamers
i agree it wasn't that smart to try over and over to pillage, it was a losing effort - with luck of course.
my question is this: i am confused because it looks like your saying an AI unit can attack-withdraw-attack-withdraw-attack-withdraw in rapid session without:
Giving your unit a chance to fire back
Passing the "combat" for you to have the option of not using the next pillage
my question is this: i am confused because it looks like your saying an AI unit can attack-withdraw-attack-withdraw-attack-withdraw in rapid session without:
Giving your unit a chance to fire back
Passing the "combat" for you to have the option of not using the next pillage
Incorrect Roland. It's mathematically intelligent for the AI sub to continue attacking.
It may only possess a 6% chance of an outright win, but it also has a 75.2% chance of retreating (.8*.94), which provides it with an 81.2% chance of surviving the battle. That 81% is the crux of the matter, and is what enables it to win. It means that it has an 81% chance of damaging the defender without taking any damage itself. And, if it actually does succeed in damaging the defender, than each subsequent attack simply increases its mathematical edge. That's not dumb, that's smart. I've said this numerous times, this is not a foolhardy attack by the AI that has no chance of succeeding. It is mathematically intelligent, successful (80% successful in fact) and common.
It may only possess a 6% chance of an outright win, but it also has a 75.2% chance of retreating (.8*.94), which provides it with an 81.2% chance of surviving the battle. That 81% is the crux of the matter, and is what enables it to win. It means that it has an 81% chance of damaging the defender without taking any damage itself. And, if it actually does succeed in damaging the defender, than each subsequent attack simply increases its mathematical edge. That's not dumb, that's smart. I've said this numerous times, this is not a foolhardy attack by the AI that has no chance of succeeding. It is mathematically intelligent, successful (80% successful in fact) and common.
Roland Johansen
Deity
Heh, you must have missed something in this thread. Something like 90% of the posts.
When a unit tries to pillage a tile near a patrol ship, then the patrol ship intervenes. If the pillaging ship retreats, then it can try to pillage again and the patrol ship will intervene again. This continues until one of the ships dies (the pillaging ship or the patrolling ship) or the pillaging ship is out of movement points. This issue is disliked by a few in this thread. I also dislike the issue, but I don't think it gives a big advantage to the pillager and I also don't think it's an advantage for AI controlled civilisations.
thadian
Kami of Awakened Dreamers
i read the first and last page.
I understand what the issue is now, and while i look forward to doing this myself..
If i have a sub on both sides patrolling, and you pillage, will both subs try to intervene?
I understand what the issue is now, and while i look forward to doing this myself..
If i have a sub on both sides patrolling, and you pillage, will both subs try to intervene?
MyOtherName
Emperor
- Joined
- Dec 7, 2004
- Messages
- 1,526
Not so! It only has a 24.5% chance of being able to deal any damage (it has to win a round of combat), and that's only because your unit was already very weak.
It has an 81% chance of not taking any damage. It's very nearly a free shot (multiple free shots).
Balderstrom
Emperor
Correct me if I am wrong, but it is my understanding that:
Thus pillaging ships should not be able to retreat.
I also believe pillage should cost 2 MV
Units retreat on the turn they would have been destroyed.
Thus pillaging ships should not be able to retreat.
I also believe pillage should cost 2 MV
MyOtherName
Emperor
- Joined
- Dec 7, 2004
- Messages
- 1,526
19% chance of death doesn't seem like nearly a free shot to me.
I just realized there's another correction -- the odds of damaging the defender and still surviving is only 20.8%.
Indeed. It's a pretty stupid gamble for the pillager. About a 20% chance of dying for only a 25% chance of doing any damage at all. And almost zero chance of killing the defender. Dumb gamble. There are easier ways to get GG points.
Snuggy is just grasping at straws to keep his "real AI cheat" argument alive.
Snuggy is just grasping at straws to keep his "real AI cheat" argument alive.
MyOtherName
Emperor
- Joined
- Dec 7, 2004
- Messages
- 1,526
If you're going to call a 19% chance of dying each attack a (nearly) "free shot", then I think it's fair for Wodan to call a 21% chance of damaging the defender without dying an (almost) "zero chance".
JujuLautre
Deity
Ok, let's do the math:
- 5 round of combat maximum (don't know if it's true, let's say it is)
- 24.5% of chance to win an attack against the (already damaged) ship
- 6% of chance to win the combat in one round (24.5%*24.5%)
- 83.8% of chance to win an attack against the (even more) damaged ship
- 80% of chance to retreat
Possible outcomes against the initial ship:
- win in one round: 6%
- 1 damage + retreat: 14.8%
- no damage + retreat: 60.4%
- death: 15.1 + 3.7 = 18.8%
total: 100% ok
Possible outcomes against the damaged ship:
- win: 83.8%
- retreat: 13%
- death: 3.2%
total: 100% ok
Let's simplify stating that if at least 1 shot has been done to the ship, it's a win
Chances to win: 20.8 + 60.4*20.8 + 60.4*60.4*20.8 + 60.4*60.4*60.4*20.8 + 60.4*60.4*60.4*60.4*20.8 (all in percentage) = 48%
Chances to lose: 18.8 + 60.4*18.8 + 60.4*60.4*18.8 + 60.4*60.4*60.4*18.8 + 60.4*60.4*60.4*60.4*18.8 = 44%
Nothing happens: 60.4*60.4*60.4*60.4*60.4 = 8%
Considering my simplification, chances to win should be lower, and both others higher.
So. 48%win/44%lose/8%retreat, just to gain a few GG points? Seems hardly worth it to me.
(Also note that the first ship was heavily damaged, which completely bias the combat)
If something is wrong, because we don't understand to maths, please enlighten us mortals
- 5 round of combat maximum (don't know if it's true, let's say it is)
- 24.5% of chance to win an attack against the (already damaged) ship
- 6% of chance to win the combat in one round (24.5%*24.5%)
- 83.8% of chance to win an attack against the (even more) damaged ship
- 80% of chance to retreat
Possible outcomes against the initial ship:
- win in one round: 6%
- 1 damage + retreat: 14.8%
- no damage + retreat: 60.4%
- death: 15.1 + 3.7 = 18.8%
total: 100% ok
Possible outcomes against the damaged ship:
- win: 83.8%
- retreat: 13%
- death: 3.2%
total: 100% ok
Let's simplify stating that if at least 1 shot has been done to the ship, it's a win
Chances to win: 20.8 + 60.4*20.8 + 60.4*60.4*20.8 + 60.4*60.4*60.4*20.8 + 60.4*60.4*60.4*60.4*20.8 (all in percentage) = 48%
Chances to lose: 18.8 + 60.4*18.8 + 60.4*60.4*18.8 + 60.4*60.4*60.4*18.8 + 60.4*60.4*60.4*60.4*18.8 = 44%
Nothing happens: 60.4*60.4*60.4*60.4*60.4 = 8%
Considering my simplification, chances to win should be lower, and both others higher.
So. 48%win/44%lose/8%retreat, just to gain a few GG points? Seems hardly worth it to me.
(Also note that the first ship was heavily damaged, which completely bias the combat)
If something is wrong, because we don't understand to maths, please enlighten us mortals
Wodan
Deity
Hey!If you're going to call a 19% chance of dying each attack a (nearly) "free shot", then I think it's fair for Wodan to call a 21% chance of damaging the defender without dying an (almost) "zero chance".
thadian
Kami of Awakened Dreamers
now, with all the math broken down, i pose one final question:
Is this system unbalanced? If replicated with undamaged units, is this consistently reliable?
If this system is NOT balanced, what should be done to change or differ the odds?
At first it looked messed up, but with Juju's post i have to agree that this is a lot of waste, i mean how many hammers is this GG costing? Please, get that GG i would rather have a good navy. At any rate, this is one more thing i suppose i will be forced to use this function myself from now on, and claim no fault in regards to suicidal pillage attempts.
Is this system unbalanced? If replicated with undamaged units, is this consistently reliable?
If this system is NOT balanced, what should be done to change or differ the odds?
At first it looked messed up, but with Juju's post i have to agree that this is a lot of waste, i mean how many hammers is this GG costing? Please, get that GG i would rather have a good navy. At any rate, this is one more thing i suppose i will be forced to use this function myself from now on, and claim no fault in regards to suicidal pillage attempts.
Wow. I think you underestimate us.
P.S. I'm back!
When did civfanatics go to this strange black background?
Jerrymander
Epistemologist
Regardless of the viability of it, it mimics real life submarine warfare.
Spoiler off-topic forum options stuff :
Similar threads
