RNG may be faulty (suspicion)

Ask questions about DraStic or discuss compatibility issues here.
000Drew000
Posts:5
Joined:Fri Oct 13, 2017 10:11 pm
RNG may be faulty (suspicion)

Post by 000Drew000 » Sun Oct 15, 2017 4:19 pm

I've suspected for a while that Drastic's RNG capabilities may be less than perfect. I tried to gather a little evidence to possibly support this suspicion. In Pokemon Soul Silver, bellesprouts are supposed to spawn either male or female at a 1:1 ratio, according to Bulbapedia. I counted bellesprout encounters and concluded a 5:2 gender ratio at the end of my experiment. Another pokemon, weedle, has a similar 1:1 ratio. I found it at an 7:2 ratio. Also, weedles can spawn with 1 of 2 abilities. 0 out of 15 caught weedles caught had the second ability option. I also can't help but notice that when one uses the move hypnosis and misses, one may find it often misses twice or even three times in a row. This doesn't happen everytime, but often enough for it to seem less than coincidental. Does anyone have any thoughts?

Fefo
Posts:263
Joined:Sat Oct 26, 2013 8:59 pm

Re: RNG may be faulty (suspicion)

Post by Fefo » Sun Oct 15, 2017 5:10 pm

Could be, but you need a test pool at least in the hundreds for some meaningful data. Never noticed anything absurd while emulating.
BE ATTITUDE FOR ENGRISH

Exophase
Posts:1716
Joined:Mon Aug 05, 2013 9:08 pm

Re: RNG may be faulty (suspicion)

Post by Exophase » Sun Oct 15, 2017 8:19 pm

DS doesn't have a hardware RNG, so the games have their own software RNGs that may be seeded with the time.

By default the emulator will resume the time at whenever a savestate was made so if you keep loading a savestate you may end up with similar outcomes. You could get around this by using the option to sync with the device's time or you can load from an in-game save.

000Drew000
Posts:5
Joined:Fri Oct 13, 2017 10:11 pm

Re: RNG may be faulty (suspicion)

Post by 000Drew000 » Sun Oct 15, 2017 8:50 pm

@Exophase. So you're saying the potential problem might exist in the rom instead of Drastic? Because if so, I might just be able to find another version of the game.

Jay Haru
Posts:1796
Joined:Sat Aug 10, 2013 1:10 pm
Location:Pasig, Philippines

Re: RNG may be faulty (suspicion)

Post by Jay Haru » Sun Oct 15, 2017 10:03 pm

No. he is saying that the issue here is tied to whether you are using an ingame or the save state regarding the game you are playing.
Behold my mighty
Image

User avatar
Platybus
Posts:40
Joined:Sat Jul 08, 2017 9:44 am

Re: RNG may be faulty (suspicion)

Post by Platybus » Mon Oct 16, 2017 4:55 pm

Hey, how's it going? It's good to see a report on a possible misbehavior, but what you said is actually related to the way Pokémon SoulSilver has been developed. Let's see:
000Drew000 wrote:In Pokemon Soul Silver, bellesprouts are supposed to spawn either male or female at a 1:1 ratio, according to Bulbapedia. I counted bellesprout encounters and concluded a 5:2 gender ratio at the end of my experiment.
Since RNG means Random Number Generator, it's supposed to be unpredictable. When you walk through the tall grass, you have a 50% chance of encountering a male Bellsprout, and a 50% chance of finding a female one. The outcome of this situation doesn't affect the next Bellsprout encounter, so, even if you're guessing you'll battle one of the two genders, the 50% chance can trigger any of these.
000Drew000 wrote:Another pokemon, weedle, has a similar 1:1 ratio. I found it at an 7:2 ratio. Also, weedles can spawn with 1 of 2 abilities. 0 out of 15 caught weedles caught had the second ability option.
Like with Bellsprout, odds apply with Weedle. Besides, in Pokémon SoulSilver (a Generation IV game), the only ability Weedle can have is Shield Dust. From Generation V onwards, Weedle has Run Away as a Hidden Ability.
000Drew000 wrote:I also can't help but notice that when one uses the move hypnosis and misses, one may find it often misses twice or even three times in a row. This doesn't happen everytime, but often enough for it to seem less than coincidental. Does anyone have any thoughts?
In SoulSilver, Hypnosis has an Accuracy of 60%. That means, the move will miss 4 out of 10 times. Have in mind your chances of hitting are lower if your rival had its evasion raised, or if your Pokémon had its accuracy lowered. You can see how the accuracy rate behaves here: https://bulbapedia.bulbagarden.net/wiki ... c#Accuracy

000Drew000
Posts:5
Joined:Fri Oct 13, 2017 10:11 pm

Re: RNG may be faulty (suspicion)

Post by 000Drew000 » Mon Oct 16, 2017 7:41 pm

You seem to be right about hypnosis being 60% (my fault). Yet that doesn't change the fact that it seems to miss in successions a lot. I don't mean after reloading from a savestate either. I mean three real turns in a row. Then again, that part of what I'm saying is more suspicion based than evidence based, so I can give that the benefit of the doubt.

As for the gender ratio, 50% chance between two outcomes is a 1:1 ratio. If I encounter a bellesprout, run from the battle, encounter another bellesprout (wothout reloading from a savestate), and repeat those steps, I should be seeing a roughly equal distribution of genders over time. If the outcome is not close to that after many trials, and after the same thing with weedles and the abilities, then it's doubtful outcomes are being generated at random equally. Please let me know if I'm still misunderstanding something.

Jay Haru
Posts:1796
Joined:Sat Aug 10, 2013 1:10 pm
Location:Pasig, Philippines

Re: RNG may be faulty (suspicion)

Post by Jay Haru » Tue Oct 17, 2017 3:53 am

from what i understood, a 50% chance does not equal to a 1:1 ratio.

a chance equation, it means you have 1 in 2 possibilities. this does not necessarily mean they would be of equal occurrence. thus fulfilling the term 'random' in rng

compare to a ratio wherein if you meet a male pokemon, that means your next encounter would surely be a female and vice versa. it would not be random anymore since the next pokemon you meet is basically set.

this goes the same for the unseemingly low/high hit rate of any skill. it only has a certain amount of chance hitting. it does not mean 6 out of 10 that it will hit/miss the target always.

think of it this way, a change encounter is like rolling a dice or flipping a coin which is basically the rng but ratios are like a pattern or code that has a sequence and is never random.
Behold my mighty
Image

000Drew000
Posts:5
Joined:Fri Oct 13, 2017 10:11 pm

Re: RNG may be faulty (suspicion)

Post by 000Drew000 » Tue Oct 17, 2017 11:25 pm

I understand that the ratio is not guaranteed to be met exactly, but in a large sample size of multiple trials, one should be seeing a roughly 1:1 ratio for 50% accuracy. Seeing something radically different is definetly possible, but statistic very improbable.

Fefo
Posts:263
Joined:Sat Oct 26, 2013 8:59 pm

Re: RNG may be faulty (suspicion)

Post by Fefo » Wed Oct 18, 2017 10:47 am

Is there any data? A spread like 113 to 87 for 200 tries would be aceptable for a 50% chance. Longer tests are needed for accuracy however, and small samples of repetition means little.

Anyway, the only game I ever heard that messed with RNG was Phantasy Star Zero, as a piracy protection. Since it's rates are brutal as is, no idea if they match what should be real.
BE ATTITUDE FOR ENGRISH

Post Reply