1. We have added a Gift Upgrades feature that allows you to gift an account upgrade to another member, just in time for the holiday season. You can see the gift option when going to the Account Upgrades screen, or on any user profile screen.
    Dismiss Notice

A "real" AI cheat

Discussion in 'Civ4 - General Discussions' started by The Snug, Jul 4, 2008.

  1. The Snug

    The Snug The Civ Heretic

    Joined:
    Dec 5, 2003
    Messages:
    954
    Gender:
    Male
    Location:
    Seattle
    Has anyone else noticed the withdrawal-blitz ability the AI has for all of it's fast attack units?

    The most notable usage is with attack subs who I've seen attack 4 or 5 times before dying. I used to think that I was being attacked by like 10 subs; eventually I realized that it was just a handful of subs attacking, withdrawing and attacking again.

    I've also noticed this with mounted units. I've seen cavalry attack, withdraw, and attack right away again.

    Then again, I don't really remember attempting to see if my units can do the same. At any rate, even if my units also possess this ability, this should be considered a game error because this is a function that is only supposed to exist for blitz promoted units.

    Editing Note: After lengthy discussion, it has now been proven that in a specific scenario, AI attack subs can attack and withdraw for as many times as it has movement points available. This scenario is when a human player assigns naval ships to protect resources by utilizing the patrol function. Any AI ship that attempts to pillage those resources may do so for as many times as it has movement points available, thus via a mechanism of the game, each attempt at pillaging will result in combat between the attacking sub and the defender.

    It has also been shown that mounted units do not possess this ability, since no patrol feature exists for land units. Although, it would be interesting if a patrol feature did exist for land units, whereby you could assign sally defenders for cottages and resources. That's a feature we need.
     
  2. JujuLautre

    JujuLautre Deity

    Joined:
    Apr 9, 2007
    Messages:
    3,112
    Location:
    Kansai, Japan
    I suspect a glitch. Do you have a save?
     
  3. Balderstrom

    Balderstrom Emperor

    Joined:
    Oct 25, 2007
    Messages:
    1,141
    Location:
    Halifax, NS, Canada
    Aren't the attack subs the units that can Launch missile attacks?
     
  4. The Snug

    The Snug The Civ Heretic

    Joined:
    Dec 5, 2003
    Messages:
    954
    Gender:
    Male
    Location:
    Seattle
    No specific save. Not a glitch, but a recurring error. It's just something that I notice on a regular basis. Pay attention to the chariot, horse archer, cavalry, knight, elephant attacks, you'll notice it.

    Indeed, in the beginning I thought I was just imagining it, or that I wasn't paying attention to the quantity of enemy units, but it has occassionally been so obvious that I can no longer deny it.

    If you are diligent in your recon, you might notice only a single one of these units in your area, but you might also notice that it attacks you twice.

    Both sub types can do it (IIRC) but mostly the attack sub (non-missile carrier) will bcz of it's 80% withdrawal capability. That's why a single attack sub can attack 4 or 5 times (bcz it simply continually withdraws). Most of the time it's not a big deal, but when your defending unit is just as beat up as the attacker, it can result in the loss of the battle rather than for an opportunity to retreat/reinforce. Especially annoying when it comes to city defense.
     
  5. MyOtherName

    MyOtherName Emperor

    Joined:
    Dec 7, 2004
    Messages:
    1,526
    In other words, you have no proof -- you simply have supreme confidence in your powers of observation and understanding of what you see, despite the fact you never actually pay close attention or investigate the situation!

    (This was not meant to be insulting -- it is only meant to be frank. It's far too easy for us to delude ourselves based on this sort of idle speculation; you can only be sure if you spend some effort investigating the situation)

    It's nearly impossible for that to happen (except for siege). When a withdrawl happens, the attacker is not merely "beat up" -- it is literally one hit from being dead. Withdrawls only happen when the defender deals what would be a killing blow, at which point the attacker gets a chance to escape instead of dying.
     
  6. Psyringe

    Psyringe Scout

    Joined:
    Dec 7, 2001
    Messages:
    3,394
    Location:
    Berlin, Germany
    I made a testbed and couldn't reproduce your observations at all.

    Download the save I attached, and just hit "End turn". You're Mansa Musa. Bismarck will attack your city. Note that one f his cavalries will withdraw from combat. It will not attack again in the same turn, contrary to your claims.

    Edit: I updated the tesbed. There's now also a submarine (which I gave +60% withdrawal chance plus the +50% it has already) which will attack your destroyer, withdraw, and (contrary to your claims) not attack again.

    Feel free to modify the testbed if you want to provide evidence. However, I doubt that you'll find one, because as someone else already has said, it would be hardly possible for a unit to continually attack and withdraw anyway, because the unit would most likely be destroyed on its second attack.
     

    Attached Files:

  7. Balderstrom

    Balderstrom Emperor

    Joined:
    Oct 25, 2007
    Messages:
    1,141
    Location:
    Halifax, NS, Canada
    It's also possible there is a whole army of subs outside your FOV, in the "grey fog of war"

    Also when enemies have insane stacks, you will no longer see each individual unit, even on the hover over,
    Cavalry (23)

    I recall more than a few times I went to go wipe up a "handful" of defenders, and there were 30+ heh.
     
  8. JujuLautre

    JujuLautre Deity

    Joined:
    Apr 9, 2007
    Messages:
    3,112
    Location:
    Kansai, Japan
    Exactly. I suspect more units are present, and that the OP is confused by the graphics he sees. Once again, a save would be most welcome

    to the OP: I don't know if you realize the number of times someone "thinks" he sees or understand something that goes against the rules, where in fact there is an obvious explanation. I'm included in this number, I already posted 2 times for bugs which were not bugs...
     
  9. The Snug

    The Snug The Civ Heretic

    Joined:
    Dec 5, 2003
    Messages:
    954
    Gender:
    Male
    Location:
    Seattle
    Condescension not required. I've probably put more hours into civ than anyone else here. I'm not anal enough to save every turn before this anomaly happens (since I cannot predict it's occurence). I am certain of my observations.

    I've seen it a number of times that when I attack a stack of units outside my city that only a single AI horse archer is left alive. On the subsequent turn, that solitary horse archer attacks twice (withdraws, and immediately attacks again). And no, a second horse archer does not appear out of nowhere to attack, for if a second horse archer had, than the original retreating horse archer would still be sitting within two squares of my city.

    I used to only play archi maps and have hundreds of hours of experience with AI attack subs and have seen this. You're remarks are frivolous. I don't think testing is conclusive since the AI isn't consistent with this ability.

    Psyringe, I'm aware of the mathematics, which is why it is more likely with attack subs than with mounted units (80% withdrawal).

    Since y'all or being such snot heads I will attempt to keep a save of this next time I see it happen.
     
  10. CoZe

    CoZe Warlord

    Joined:
    Oct 19, 2007
    Messages:
    193
    don't forget you have auto saves. next time this happens, just roll back to nearest autosave, do everything as exactly you did before and save. it should happen (unless you play with random seed, which is off by default).
     
  11. DanF5771

    DanF5771 Emperor

    Joined:
    Feb 21, 2008
    Messages:
    1,194
    Sirsnuggles: Do you then get multiple messages of "unit X has withdrawn from combat with unit Y"?
    Do these "fights" really cost you HPs of your defenders? Maybe it's just a graphical problem with repeated combat animations.
     
  12. JujuLautre

    JujuLautre Deity

    Joined:
    Apr 9, 2007
    Messages:
    3,112
    Location:
    Kansai, Japan
    I don't care if you think I'm condescendent or what, it's your problem if you don't want to believe that I just want to help.

    However, I won't accept that you call everyone here snot heads when we're just trying to understand what's happening and solve your problem.

    That said, a save would be most welcome. Thanks.
     
  13. The Snug

    The Snug The Civ Heretic

    Joined:
    Dec 5, 2003
    Messages:
    954
    Gender:
    Male
    Location:
    Seattle
    I do get multiple messages.

    I suppose the next time I notice it, I could immediately create a save that would show the combat calculator thingy.
     
  14. Woody1

    Woody1 Prince

    Joined:
    Feb 19, 2002
    Messages:
    468
    Location:
    Texas
    I've never noticed anything odd, myself. I firmly believe the OP is just another "the AI cheats at combat!" screamer. Every single time, they're proved wrong. Yet, they keep on claiming it again and again.
     
  15. reverend oats

    reverend oats King

    Joined:
    Dec 18, 2007
    Messages:
    641
    Location:
    Uberwald
    Sigh.
    THE AI DOES NOT CHEAT IN COMBAT!
     
  16. onedreamer

    onedreamer Dragon

    Joined:
    Oct 21, 2004
    Messages:
    6,580
    Location:
    Torino - Italia
    I'd rather complain about the (real) fact that the AI does withdraw even with ridiculous chances while the human player must make a big sacrifice to his/her God in order for this to happen just once !
     
  17. Psyringe

    Psyringe Scout

    Joined:
    Dec 7, 2001
    Messages:
    3,394
    Location:
    Berlin, Germany
    No one's been condescending, SirSnuggles. The situation we see is that you claim to have made an observation ...
    a) that no one else has made
    b) that hasn't been reproducible at all when we tried to
    c) that is in stark contradiction to known combat mechanics (i.e., a unit that did what you claim to have observed would most likely die on its second attack and never be able to even try any subsequent attacks)
    d) that isn't backed up by the combat code (which is exposed in the DLL source code)

    The evidence you have provided so far is that you ...
    a) maintain that your observation must be correct
    b) claim that you put more time in Civ than anyone else

    Based on the facts we have, we chose not to believe your claim that the AI cheats by re-attacking with a non-blitz unit in the same turn. You may call that snotty and frivolous, but imho it's rather just common sense, really. Look at the facts on the table and decide yourself.

    Btw, here's another testbed for your claim. The following save has 60 combat situations of 2 attack submarines against one destroyer and two galleys. All attack submarines have maximal withdrawal chance. Combat takes place in enclosed areas to make sure that other units don't disturb it. You'll see lots of the following results:

    a) first sub kills destroyer, second sub kills galley, no subsequent attacks from either sub
    b) first sub withdraws from destroyer, second sub kills destroyer, no subsequent attacks from either sub
    c) first sub withdraws from destroyer, second sub withdraws from destroyer, no subsequent attacks from either sub

    The only result that never ever happens is the one you claim to observe:
    d) sub withdraws from combat, then attacks again

    It simply doesn't happen. As I said before, feel free to modify the testbed in any way to see fit to back up your claim, but I remain highly skeptical. Also, of course, a save from just before such a combat (*not* afterwards) would serve to back up your claim.

    Another thing you should do is checking your CustomAssets. There is a (slight) chance that the observations you claim to have made are true, but happened due to a mod acting up, which would also explain why no one else is able to reproduce them.
     

    Attached Files:

  18. Turquoiside

    Turquoiside Emperor

    Joined:
    Jun 18, 2008
    Messages:
    1,552
    I have seen an AI cheat in my current game. Alex is one a one tile Island and as of screenshot , he can't trade with anyone. So how the heck does he have Swordsmen?!? Unless he got really lucky and had iron one that one spot, he dosen't! You can see the shot in my image.

    P.S. Though he is cheating himmself, he has had that GS for about 20 turns now.
     
  19. Psyringe

    Psyringe Scout

    Joined:
    Dec 7, 2001
    Messages:
    3,394
    Location:
    Berlin, Germany
    Well ... and where's the evidence to back up that claim?

    Again, this claim can be tested quite easily. Here's a testbed, a battle with 500 attack subs on each side. Each attack sub has a withdrawal chance of 50%, no other bonuses are involved. Hence, if you attack your opponent's stack with your stack, the following results are to be expected:

    - Since each battle has a 50% chance to be won, you should (on average) win 250 battles, and lose 250 battles, leaving Bismarck with 250 subs afterwards.

    - From the 250 battles you lost, half of the time your sub should be able to survive, leaving you with 375 subs aftwards.

    Now, load the save I attached. Make sure to set the options for offensive and defensive quick combat, and stack attack. Then form a stack with your 500 subs, attack Bismarck's subs, wait a while while the PC is calculating (takes about 4 minutes on my machine), then check the results. You can see the number of your units that have survived in the lower left. You can see Bismarck's remaining units if you mouse over his stack and add the number of subs that are listed individually (usually 15) to the number of subs that is displayed as a single entry (bottom of the list, number in parantheses).

    You'll see that Bismarck has 242 subs remaining, this matches the expected value very closely (actually much closer than expected, you'll always get *some* random deviation). So, you've lost 242 combats. Half of the time your sub should have been able to retreat, so you should have 258 (number of won combats) + 121 (number of successful retreats) = 379 subs remaining. Actually, you have 382 subs remaining - again, very close to the expected number.

    Now, reload the save and do not attack, skip turn for all of your subs. In the inter-turn, Bismarck will attack you with his stack. Wait until he's finished, then check the numbers.

    You'll see that you've got 237 subs left (13 less than expected, but such a deviation is well within reason for random combats). This means that Bismarck should have 263 (number of won combats) + 118.5 (number of successful retreats) = 381.5 subs. Actually, he has 389 subs left, again very close to the expected number.

    So, where is your "real" fact that the AI has better withdrawal chances than the human player?
     

    Attached Files:

  20. Psyringe

    Psyringe Scout

    Joined:
    Dec 7, 2001
    Messages:
    3,394
    Location:
    Berlin, Germany
    You forgot to attach the screenshot. ;)

    Please upload the savegame, so that we can check it out. A savegame from a turn before Alexander required iron would be helpful as well. Btw, how did you manage to get an AI to start on a 1 tile island? I've never seen this.
     

Share This Page