The random character select algorithm


#1

I’ve recently started using our new friend the [?] box quite freqently… pretty well all the time. I had no idea it could be so much fun!

However, I’ve been through, ohhh, I’m going to guess about 100 games using it now… and I’m getting the distinct feeling that it’s not actually very random; I get certain characters wayyyy more often than others. In a perfect world, I’d have seen each of the seventeen character about five times, six for a few of them. My experience hasn’t even been close to that. At all.

Here’s a quick scatter of how the choices have turned out, ranked from most frequent to least frequent.

  • Very common:
    Dhalsim
    Vega
    Ken

  • Common:
    Bison
    Sagat
    Balrog
    Honda

  • Occasional:
    Blanka
    Chun
    Cammy

  • Rare:
    Guile
    Hawk

  • Once:
    Deejay
    Gief

  • Never:
    Ryu
    Akuma

Take this with a grain of salt. Keep in mind that these are all estimations and only to the best of my recollection. I know my sample group is a little on the small side to accurately test how close it is to “true” random, but I strongly feel like I’ve been getting some pretty lopsided results. (For the record, I’m on PS3, and this has been mostly for Scoreboard Quick Matches, some Friendly Matches.)

Has anyone else been using Random Select often? Please add and compare your results with mine.

If the algorithm is not drawing randoms properly and there’s talk of patching SFHD, I’ll make a Capcom Unity post in hopes of getting this shit straightened out. Heck, maybe I should just file a bug report.


#2

ive never got akuma with random.

ive got hawk and gief once.

dee jay 4 times

guile, ryu balrog and blanka a trillion times.


#3

I doubt they messed up the random select. If you know anything about computers “random” is about as easy as it gets (well true random is as hard as it gets but that’s another story, ps3 random is as random as you will ever need for 17 possible outcomes).


#4

If you really want to find out how random it is, you need a much higher sample size, and keep track of who gets picked for you.


#5

Thanks for the responses, guys. :smile:

YieArKungFu: Yeah, very good point and I know. True random is impossible. Decent approximation random-enough-for-all-intents-and-purposes is easy enough though, and I don’t doubt that the PS3 has some stock methods built-in for developers, so it seemed odd that it would be screwed up.

CWheezy: Agreed and previously acknowledged. This is why I made the thread, to gather more results! :bgrin:

I’ll throw some (addmittedly extremely) guessy numbers on these so people can at least get a better rough idea of how biased it’s looked to me thus far.

Dhalsim 16
Vega 14
Ken 12
Bison 11
Sagat 10
9 Balrog
8 Honda
5 Blanka
4 Chun
4 Cammy
3 Guile
2 Hawk
1 Fei
1 Deejay
1 Gief
0 Ryu
0 Akuma

(I guess forgot Fei in the original list.)

Of course it’s always 100% possible that a particular given outcome could happen under a perfect random algorithm, but again, of course the odds are stacked against this.

I’m gonna keep a notepad by my stick and start tracking this stuff so I can gather some hard statistical evidence. I’m not trying to cry wolf here, I hope the random works fine too. It just doesn’t feel like it yet.


#6

there are two ways they could have gone about random select ;

  1. Pick a random # from 1- 17 then move the curser around in a semi random looking pattern to get there or have the ? box cycle through portraits and stop on that one or whatever

  2. move around the board randomly and stop at a random time.

If the results are not coming out random, then they picked number two. This means whatever randomizing pattern they are using rarely passes certain characters and doesnt go to Akuma because he is off screen.


#7

#2 definitely sounds like a stupid way of doing it, but it’s certainly worth mentioning that I don’t think the portraits cycle through Akuma. I’ll have to look more closely.

That doesn’t necessarily mean anything though, the portraits could just be a simple animated loop that actually holds no meaning. (It would just change to the portrait of the randomly selected character on the frame after your button press.) I actually had always just assumed it would be that, just a visual flair rather than an actual roulette going on.

I think I’ll just make a thread about it on Unity. I’m sure Rey or someone would be able to answer some of these questions for us, and save us all the silly conjecture.


#8

Akumas portrait has shown up for me on the ? box but ive never got him with random.

I played some more tonight with random and ended up getting rog and deejay like constantly. Honda is pretty common too.

gief, hawk, chun, and fei are really rare for me though


#9

Wow, your results are quite different from mine. I’ve played about 100 matches over the past 2 days against a friend, using random all the time. I tend to get Claw and Honda A LOT. I have NEVER gotten Dee Jay, Fei Long, or Akuma. I’ve only gotten Dictator once, which sucks, because he’s my main. :sad:


#10

^ Mmm, it’s great your results are so different from mine and those of gridman. Was that on 360 or PS3? (Probably doesn’t matter but trying to be thorough.)


#11

It was on PS3. Heh, I have a live account, but presently I don’t have a 360. If I had a 360 and a good stick, I’d be on there in a heart beat just because more people have voice chat, and it seems like there is better comp there. :sad:

As far as tracking results, it should be easy to do with HD remix. The versus screen has a nice record of the matches for easy counting.


#12

you’re sample size is way too small for 17 characters.


#13

My results are also different, but don’t feel “random” exactly. Lots of Honda, lots of Bison, lots of Fei Long (I once got Fei Long six games in a row). A friend of mine got lots of Cammy, which I’ve never gotten with Random. It seems very “streaky.”


#14

Akuma will never show up on random, at least not from any random anybody has ever done: however, when you come back from a match and your cursor is on Random, Akuma is the first portrait shown for a split second. Really weird.


#15

EVERYONE LOOK, it’s a master-genius statistician. Thanks for stopping by! :tup:


#16

Actual randomness is not going to feel random in any intuitive human fashion, meaning you won’t get anything that seems like an even spread of character picks. People tend to intuitively think random means “different all the time”, and their brains pick up on unexpected patterns like seeing Guile more than once in a few picks, or twice or three times in succession, and remember and accentuate that experience.

I think the only character I never get in random is Akuma. Today I got Zangief 3 times in 5 picks. But that’s randomness for ya.

It’d be kinda interesting if one random box was normal random, and one was a learning random, biased towards previously unpicked or infrequently picked characters. More of a “variety” box than a random box.


#17

If it is using my second option or something similar, you will likely get different results from the right and left random selects. If people are going to compare the results of the random select you should separate results from both sides. Akuma wont show up much if at all because he can only be moved onto by the randomizer from one portrait (the one bellow him) as opposed to the 4 possible paths for the other characters (-1 for each character beside the random select squares) Ryu, Boxer, Guile and Sagat. Also, I have not had a chance to play this game yet. If the select cursor doesn’t wrap around from one side to the other that will effect how many possible ways the select cursor can land on that character.

The option 2 I mentioned is definately a very retarded way of randomizing, and ironically is one of the more common methods employed in games for some reason.


#18

RoboKrikit: You’re absolutely right. Don’t worry, I am aware that the human mind is built to naturally, automatically try and find trends, and especially at this point I don’t doubt that I’m subconsciously looking to find certain things more than others. (“SEE I JUST GOT A VEGA AGAIN.” :rofl:)

I have actually started collecting real data. I’m jotting down every single pick I get, and I’ll happily add in the results of anyone else who’d like to contribute.

Has anyone had the random select give them Akuma or Ryu? Even once?

I mean I get if they deliberately left Akuma out of it, as lame as it may be, but the exclusion of Ryu would most definitely be an error. How evil would that be, omitting the most popular character!


#19

What method are you using to collect your data?

Always the left random select? Always the right?
Do you always try to keep the time between highlighting random-select and selecting random-select constant?
Do you always use the same button to select random?
etc


#20

Not that this necessarily means anything for STHD, but in CvS2 at least, the character that’s being displayed is the character that gets picked. You can actually have more than one of the same character on a team this way.