Judgment Hammer: anything unknown?

[ant-d]

Do we know everything about the hammer (side-B)?

Have the "rules" been tested? Supposedly x,y,x and opening with 1,2 is impossible (without dying).

Have the occurrence rates been tested?
Supposedly around 1/7 chance of getting x.

And how about the PRNG and the nature of the randomness?

By "tested" I mean rigorous controlled tests using large sample sizes (at least a few thousand hammers). The code has probably been reversed by now, but it would be interesting to see the real-world stats.

 

[bubbaking]

I think all of these traits have been tested, although a sample size of a few thousand might be a little extraneous...

 

[Oracle]

you have tas melee right? Supposedly if you hammer in the exact same situation multiple times you'll get the same result. Just make a save state and repeat and see what happens.

If it turns out it is situational, then the hammer being random is disproved so we won't need a huge sample sized test

 

[bubbaking]

Oh, I don't have TAS Melee, but people already have tried the hammer with TAS. The RNG in Melee is not truly random, so thus, nothing in the game is truly random either. However, so many things, including stuff outside of the match, go into deciding the number that it may as well be random. However, if you save a state of the game and perform the hammer at the exact same frame each time, you will always get the same number.

 

[Oracle]

I was actually talking to ant-d because i'm pretty sure he has it. Although that theory is predominant i haven't actually seen any tests or evidence so i'm still skeptical

 

[100 proof%]

But at what point are they made random? I've done tests with peach by doing slow-mo matches and at the start of each match on FD I got like too many different turnips for it to be random. I'd think watch would be less random. But like when does it become random? Like maybe the 4th frame makes it set.

 

[bubbaking]

Slow-mo doesn't cut it. You need something that can slow down the game to a frame-by-frame basis. The turnips and hammer aren't made 'random' at any point. If you start a match with G&W on Final D with the same exact conditions and actions taken before the match, then on the first frame possible, G&W will always get the same hammer number. This is only testable through TAS, cuz I think the GCN's inner clock may actually influence the RNG.

 

[Celestia]

opening with 1,2 is impossible

really? g&w is my favorite character and i've played him for a really long time, and i could have sworn that i've opened with a 1,2 WITHOUT dying. strange.

 

[bubbaking]

I'm pretty sure it's not possible. However, consider this: if you input side-B while G&W is not busy doing another action, it will register, even if it never actually comes out. For instance, if you side-B and a 7 was about to come out but then you're hit on the very first frame of startup, the 7 still registers so you can't get it for another two hammers. This could be how you 'opened' with a 1 or a 2 before you lost a stock; you probably had other failed hammer attempts first.

 

[Celestia]

I'm pretty sure it's not possible. However, consider this: if you input side-B while G&W is not busy doing another action, it will register, even if it never actually comes out. For instance, if you side-B and a 7 was about to come out but then you're hit on the very first frame of startup, the 7 still registers so you can't get it for another two hammers. This could be how you 'opened' with a 1 or a 2 before you lost a stock; you probably had other failed hammer attempts first.

hey, i think you're right! good explanation. yeah that makes sense.

 

[bubbaking]

Btw, welcome to Smashboards!

 

[Celestia]

Btw, welcome to Smashboards!

haha thanks

 

[ant-d]

Pending Judgment hammer dataset

I now have a dataset consisting of tens of thousands of judgment hammers.
The script is still running. I'll create a new topic with results and analysis once I have a large enough sample.

24 hours of judgment hammering should be enough. Currently averaging just under 1 per second.