RNG nuances

[Irysa]

So I've been doing some testing with the RNG with Bookmark backups (I should really just be using Casual to test this stuff but w/e, I have a good scenario on my run).

The setup is in P-4 where I have Sumia about to level up and a wide variety of different things I can do to tinker with RNG. So far, by standard expectations of what burns an RN, I can simply count RNs burnt and predict the level up if I've got a precent for previous burns, even if I use different units and different triggers. If you're familiar with the RNG in the series then you'll be familiar with most of these;

1. Burn 2 RN and average them first to check hit
2. If hit is calculated to be successful, burn 1 RN to check for each applicable skill activation in skill priority order, and 1 RN for crit.
3. If enemy is not dead, burn 1 RN for Dual Strike if applicable (does not burn the RN if chance is 0%, ie no support unit or unequipped weapon on partner), and if successful, preform all the same RN burns for the regular attacks.
4. Enemy counterattack uses the same formula, but Dual Guard burns 1 extra RN and IT IS BURNT BEFORE HIT CALCULATION IS EVEN DONE, so this is burnt no matter what if you've got a support unit.

If I burn a total of 12 RN's through 2 instances of combat, here are three seperate examples that get the same level ups, so I can assume the methodolgy to be correct.

Lon'qu + Unequipped Robin attack Barbarian. (3 RNs) Barbarian counterattacks and misses. (3 RNs) Lon'qu attacks and KO's Barbarian. (3 RNs) Sumia KO's Myrmidon in one hit. (3 RNs) = 12 RNsLevel up for HP LCK RES

Libra + unequipped Gaius double attack Archer. (6 RNs) Sumia + unequipped Chrom double attacks Myrmidon at 2 range. (6 RNs) = 12 RNs Level up for HP LCK RES

Gregor (armsthrift) + Gaius attack Myrmidon. (5 RNs) Myrmidon counterattacks and hits. (4 RNs) Sumia KO's Myrmidon in one hit. (3 RNs) = 12 RNs Level up for HP LCK RES

Now this is where I am slightly unclear because there is a factor that seems to mess things up, and that is the amount of instances of combat entered.

Lissa KO's Archer in one hit. (3 RNs) Cordelia KO's Myrmidon in one hit. (3 RNs) Sumia + unequipped Chrom double attacks Myrmidon at 2 Range (6 RNs) = 12 RNs Level up for HP SPD LCK DEF

So I'm thinking that for each instance of combat, burn one extra RN? I have other smaller examples that have disrepancies too, so this seem to be the easist assumption I can make about the RNG (Calc 6 RNs for Sumia to KO Myrm with unequipped Chrom at 2 range, Calc 6 RNs for Lissa to KO Archer then Sumia KO Myrm in one hit, level ups are different.)

I expect shadowofchaos to probably be the most well informed on this, but I may as well ask to be sure.

Edited September 18, 2014 by Irysa

[Czar_Yoshi]

Nobody's ever done an extensive study on exactly how Awakening's RNG works, so you're not going to get anything conclusive from anyone else. I'm certainly very interested, though.

So, are you sure that every battle doesn't use an RN for every possible option? Since you get the same results with two battles every time but a different one with three battles... To check that, you should do a two battle set with one extra (according to your method) RN and see if the level up is the same as the 3 battle set.

Edited September 18, 2014 by Czar_Yoshi

[Irysa]

I'm pretty sure it doesn't use an RN for every possible outcome because if it did, there would be an inconsistency in the numbers for scenarios where there are misses. Not to mention I've done tests where the target is dead before a DS can actually occur (skill), and the results match up for burning 4 RNs instead of 5. I've also done testing for 3 and 4 battles and can get consistent predicted results in those scenarios too.

However, I did try out what you said and dissapointingly it doesn't seem to add up, indicating that it's more complicated than simply adding on an extra RN. :\

I actually went up to 18 RNs in two battles and could not get another HP SPD LCK DEF to occur. There are still some more things I could try but it does appear that extra battles throw a complete wrench in the gears.

Edited September 18, 2014 by Irysa

[XeKr]

afaik the prng of post gba games is more sophisticated, such that it's difficult to manipulate by hand (possible some time component or w/e). FE9 TAS suggests rng is controllable there for example, and I've noticed a bit of it with save states, but it's not really feasible in conventional playthroughs. Moreover, each hard chapter reset further shakes things up. FE13 probably isn't too different.

[Tables]

The same attack can have a different animation in Awakening, right? I would expect then 1 RN is used to determine which animation is used.

If this is correct, then burning 1 more RN on the two attack test should equal things up.

[Elieson]

Does a skill procs RN actually process, if no skill exists on the unit in battle?

Edited September 18, 2014 by Sara.

[Irysa]

No, you need to have a skill for an RN burn to occur on that.

However, I did figure out something else that kind of messes everything up even more.

Fred will always activate Luna if he goes first, and Robin always activates Ignis if she goes first. Luna is Skill% chance, Fred has 15 Skill +2 from Panne so 17%. Robin has Ignis, and 16+2 Skill so 18%.

HOWEVER, Gregor's Armsthrift, which should be Luck*2% chance for a 22% chance (10 luck +1 from Cordelia) does not activate if he goes first. However the same amount of RNs appear to be being used for this as for the others, since I get the same level ups between the three.

The same attack can have a different animation in Awakening, right? I would expect then 1 RN is used to determine which animation is used.

If this is correct, then burning 1 more RN on the two attack test should equal things up.

Only just woke up but from the stuff I have atm it doesn't seem to work that way either. None of these seem to correlate to the 3 battle scenario level up.

6 > HP STR SKL SPD LCK
7 > HP SKL SPD LCK RES
8 > STR SKL SPD DEF
9 > HP STR SKL LCK
10 > HP STR SPD RES
11 > HP STR SKL DEF
12 > HP LCK RES
13 > STR SPD DEF RES
14 > HP SKL LCK
15 > MAG SPD LCK RES
16 > STR SKL SPD LCK DEF
17 > HP MAG SKL SPD LCK RES
18 > STR SKL SPD LCK RES
19 > HP STR SKL SPD LCK
20 > HP SKL SPD LCK

Edited September 18, 2014 by Irysa

[Czar_Yoshi]

afaik the prng of post gba games is more sophisticated, such that it's difficult to manipulate by hand (possible some time component or w/e). FE9 TAS suggests rng is controllable there for example, and I've noticed a bit of it with save states, but it's not really feasible in conventional playthroughs. Moreover, each hard chapter reset further shakes things up. FE13 probably isn't too different.

Time doesn't factor into Awakening's RNG. If it did, you could change the RNG just by waiting, and we know that isn't possible.

Fred will always activate Luna if he goes first, and Robin always activates Ignis if she goes first. Luna is Skill% chance, Fred has 15 Skill +2 from Panne so 17%. Robin has Ignis, and 16+2 Skill so 18%.

HOWEVER, Gregor's Armsthrift, which should be Luck*2% chance for a 22% chance (10 luck +1 from Cordelia) does not activate if he goes first. However the same amount of RNs appear to be being used for this as for the others, since I get the same level ups between the three.

AT is probably checked somewhere else in the line from the other procs, which makes sense as it doesn't act like others (can overlap with other procs, trigger in the back, and doesn't have an activation cutscene).

[dondon151]

Time doesn't factor into Awakening's RNG. If it did, you could change the RNG just by waiting, and we know that isn't possible.

you can't change the RNG in FE11 or FE12 just by waiting either, at least not in the middle of a chapter. unless i'm misinterpreting you, i'm pretty sure that the RNG is seeded the moment the game powers on, and the seed is based on the clock.

[Czar_Yoshi]

Well, I was specifically talking about Awakening's RNG (haven't messed with Tellius's at all) but I'm pretty sure you're right.

[Irysa]

you can't change the RNG in FE11 or FE12 just by waiting either, at least not in the middle of a chapter. unless i'm misinterpreting you, i'm pretty sure that the RNG is seeded the moment the game powers on, and the seed is based on the clock.

11 and 12 have some really weird nuances in how button inputs affect the RNG, like canceling animations and stuff at specific timings appears to manipulate level ups somewhat. I messed around with this before but couldn't figure it out since it seemed almost frame specific and I couldn't be bothered to frameadvance/count frames.

Figured out some ways to get to theoretical lower numbers via some silly rescuepassing around but still no dice. If I can somehow find another HP SPD LCK DEF in two battles then I'll be a step closer to figuring this out. I'm beginning to think that perhaps the RNG just flat out advances a large amount for every battle, perhaps it uses RNs for battle quotes and animations as well?

6 > HP STR SKL SPD LCK

7 > HP SKL SPD LCK RES

8 > STR SKL SPD DEF

9 > HP STR SKL LCK

10 > HP STR SPD RES

11 > HP STR SKL DEF

12 > HP LCK RES

13 > STR SPD DEF RES

14 > HP SKL LCK

15 > MAG SPD LCK RES

16 > STR SKL SPD LCK DEF

17 > HP MAG SKL SPD LCK RES

18 > STR SKL SPD LCK RES

19 > HP STR SKL SPD LCK

20 > HP SKL SPD LCK

Going to try to aim higher now.

Edited September 18, 2014 by Irysa

[Czar_Yoshi]

You know what would be really useful for this?

A unit with equal growths in every stat. Then you could more accurately measure how many extra RNs you burn.

[XeKr]

Well, I was specifically talking about Awakening's RNG (haven't messed with Tellius's at all) but I'm pretty sure you're right.

Well wrt to the time thingy I was specifically thinking of some anecdote where iirc Vykan observed strangely consistent results (despite low probability) in his (old) FE10 run, even from hard resets, which was potentially attributable to timing the start screen. So some kind of seeding as mentioned.

But I haven't actually followed further developments much.

Edited September 18, 2014 by XeKr

[Irysa]

Three battles. Can't really discern clearly results higher than this at the minute because there are too many low RNs for crit lying around on the 9th and 10th values that make it difficult to get the desired results but I basically give up. I can't find any casual correlations between these, so whilst I can be sure that the basic stuff I outlined adds up since I can determine results within the same amount of battles reliably, I can't figure out the amount of advances used per battle. It appears to be really large so there's a chance it even advances based on another RN.

9 > HP MAG SKL LCK RES
10 > HP STR MAG SPD DEF
11 > HP STR SKL LCK RES
12 > HP SPD LCK DEF
13 > SKL SPD LCK
14 > HP MAG SKL SPD RES
15 > STR SKL DEF RES

I think to properly test this I'm going to need to have a setup later that has a variety of level ups via rescue staff abuse as well.

Edited September 18, 2014 by Irysa