This is a split board - You can return to the Split List for other boards.

random number generator and thief knife

#1cesultPosted 6/3/2009 4:55:48 PM
Hi all,
So I'm trying to steal an Economizer from Aquila (level 49) with Celes (level 6) and a thief knife. According to the algorithm, there should be ~7% of stealing something, and 1/8 chance for it to be the rare item. Out of >200 steal attempts, I got 5 phoenix downs and no economizer. Does the random number generator make it impossible to get the economizer with the thief knife, or should I keep trying? I've read the RNG isn't too random, and can make things like certain monster formations impossible. Anybody know how random it is for stealing?
Thanks,
-Ces
#2VXPoisonGasPosted 6/3/2009 5:00:42 PM
Are you using savestates?
#3cesult(Topic Creator)Posted 6/3/2009 5:23:51 PM
Yes and no. I don't reload after failing; I only load right before getting a phoenix down. Either way, in the past I've done save states in battle, and been able to get the rare item after failing or getting the common item, although this was with a high level Locke with a sneak ring. Thus I suspect the RNG to determine if the thief knife steals may be interfering with the RNG for the rare item.
#4SuperSaiyanTienPosted 6/3/2009 8:39:48 PM
IIRC,

the RNG depends on the actions you do, so if you simply "Reload" right before the message: "Stole a Phoenix Down" appears, then that won't work.

You'll actually have to change the actions taken in the battle before hand, like use an item or something, that SHOULD change the RNG.
---
~National Gamer Series~ For playstation 3 owners,
Email Darkspector31@yahoo.com for more info!
#5aerinonPosted 6/4/2009 1:26:58 PM
We're talking about 0.875% chance here (combining the 7% and 1/8). That's pretty rare. But you should get one (on average) about every 114 or 115 tries. But I guessing the standard deviation is huge on that low probability. You should get about 7 phoenix downs for each economizer. Sounds like you're actually having pretty bad luck to get only 5 phoenix downs out of 200 steals (average would be more around 14). Save state reloading and not changing actions can increase the failure rate.

Keep trying, I guess? Seems kind of masochistic to me.
#6MeepleLardiclePosted 6/4/2009 3:30:26 PM
There are some cases where the game's RNG makes it mathematically impossible to get the item. I dunno if Thief Knife for rare items at levels that low is one of those cases; I just know that getting certain items via Metamorph from a Magicite *IS* flat out impossible.
---
www.rpgdl.com
The site Meeple argues at for too much of his life!
#7livinlarav2Posted 6/4/2009 8:37:33 PM
if i remember right, you should load from before you even press Fight or Steal.

---
Mods are like cops, always there when you're trying to have a good time, never there when you need them :)
#8cesult(Topic Creator)Posted 6/4/2009 9:08:21 PM
>"There are some cases where the game's RNG makes it mathematically impossible to get the item. I dunno if Thief Knife for rare items at levels that low is one of those cases"

That's exactly what I am hoping to confirm, or hopefully, show is not the case.

>"I just know that getting certain items via Metamorph from a Magicite *IS* flat out impossible."

Thanks for this fact, I've been trying to metamorph a Cactrot using a Magicite and gave up after 10 metamorph misses.

>"if i remember right, you should load from before you even press Fight or Steal."

I've managed to steal a Debilitator from Air Force after save stating right before confirming the steal, after selecting the target. It did take a lot more tries than 1/8 chance compared to the common item, but I believe the game does not determine the random number until after executing the action.
#9ImzogelmoPosted 6/5/2009 12:58:37 AM
I have the RNG table and Excel; I can try to figure this out tomorrow if I think about it.
---
Lots of discussion of FFVI's mechanics and bugs, check out Mnrogar's Den at
http://mnrogar.slickproductions.org/phpBB3/index.php
#10aerinonPosted 6/5/2009 11:56:32 AM
I bothered to look up the standard deviation on the geometric distro (number of tries before one success) and the is huge - around 114. To extremely confident of success, I'd usually 2 or 3 standard deviations beyond the mean which translates into 342-456 tries.

Might be worth it if Imzogelmo can figure it out if there's a an RNG impossibility instead of try another 200 times. Statsitics only work on true RNGs anyway.