Advertisement
 

Membership

 

Advertisement

 

Search

 
Forums \ StarCraft forum \ Are start locations truly random?
1, 2 
Information
2 months ago

Psycho_Gemni
Are start locations truly random? (819 reads, 33 replies)
So about a month ago, I was playing vs a random guy on the ICCUP server. Well, I hosted a standard 'One on One' Python map, he chose Zerg and I chose Terran. The game started and I spawned at the 6 o'clock position. As I was building my first supply depot behind my mineral line, his overlord came into my field of vision from the 9 position, so I canceled my supply and decided to 9rax instead. I bunker rushed his 12hatch build and destroyed his natural, and he demanded a re.
We remade on the same map and once again I spawned at the 6 position and soon after I discovered that he spawned at the 9 position again. Once again I canceled my supply and killed him with a bunker rush and he demanded another rematch.
On the third game, I once AGAIN spawned at the 6 position and I don't even think I bothered to build my supply first. He was at 9 AGAIN and I won the third game in a row via bunker rushing...

Now, I believe there are 12 possible starting-position combination's for both players, so if the start locations were truly random, the chance of us spawning at identical spawn locations three games in a row is 1 / (12 x 12 x 12) = 1/1728 = ~.000579. Not too likely to happen, I would think.

Facts about the 3 games:
I hosted the games each time
We played on the same map each time
We were the same race each time
We played on One on One each time

Could these similarities have affected the chance of this happening?

This is not the only anomaly that has seemed to happen, either. If I ever play two games in a row with the same starting location in a row, I always scout the base where the opponent spawned last time, and I feel like he is there more often than he should be (i.e., he is more than 33% likely to spawn there).

Aside from this one happening and this one hunch, however, I have no other data to prove that it isn't really randomized.

So my question to GGNET is, are start locations truly random? Or are they less random given other factors?

EDIT: Well, it turns out I did my math wrong. There was a 1/144 possibility that we'd get the same spawn location three times in a row. That's a lot more reasonable than I previously thought, but if you read the posts below, you'll still probably be convinced that Starcraft's randomness generators definitely aren't as random as they should be.
Quote  
 

2 months ago
#1

LML

 (Forum moderator)   18
that's pretty usual for BW, that if you play same map/guy you happen to get 3times the same startlocation, both.
It happened to me several times o,o
Top of the page   Quote  

2 months ago
#2

gNs.I-Ahzz

   6
situations like these just stick out. you simply dont remember the normal cases where races/start locations were indeed truly random.
Top of the page   Quote  

2 months ago
#3

ETBrooD

   2
Something must be right about that theory because computers don't have a real random system. They calculate and pick locations by certain factors that I don't know. I have no idea how it works but I hope someone discovers a few facts.
Top of the page   Quote  

2 months ago
#4

ostojiy

   3
its pretty effin close to random.
Top of the page   Quote  

2 months ago
#5

080485

   11
#3 Coding that pops out random numbers is very very VERY easy to make.. i remember around 10 years back from now where i first learned to do so..its like the first step they teached to ppl back then..in a normal computer teaching class.

So i think blizzards genius masterminds have discovered that one pretty long time ago..dont you think?

Ofcourse the programs from teaching to actual game coding are very different..but it doesnt change the fact that its easy to do.
Top of the page   Quote  

2 months ago
#6

DD-Marco
interesting stuff... imagine if you knew your opponents location every time. almost like a hack :D
Top of the page   Quote  

2 months ago
#7

KawaiiRice

   1
3 bunker rushes in a row? What are you, Boxer?
Top of the page   Quote  

2 months ago
#8

makin

   3
i would like to see reps of that ._.;;
Top of the page   Quote  

2 months ago
#9

topturkey

   1
yea it happens, i was on andromeda vs a terran, i scout him first time when i was bottom left and he was top left. i go 2 gate goon and he's done.

i do it second time aswell.

and the 3rd time i do it again but he did a 2 factory noob rush so it was a bit harder but i stuck with my 2 gate reaver.
Top of the page   Quote  

2 months ago
#10

Psycho_Gemni

   1
#1 if it happens all the time, do you believe that this is one of the reasons why in professional matches etc. we do not play on the same map more than one time in a row? And do you use this to determine what your build orders are // where you scout during some games?

#3 your idea seems very skeptical, do you know precisely how accurate the current random number generators are? They can't be that far off from complete randomness if at all, can they?

#7 you had to bring that up didn't you ;___;

#8 I really don't think you do. We were just two D-ranks, lol.
Top of the page   Quote  

2 months ago
#11

ETBrooD

   2
#5 A computer has to decide which locations the players will get. So far so true, right? The conclusion would be that it cannot be random at all.
I'll explain to you what "random" really means: you have a set of cards that have numbers written on them, you turn them around so you can't see them, you look away from the cards, somebody else shuffles them, you turn back and then you choose one of the cards. That's random.
If you want this to work on a computer you must have at least 2 different programs running which isn't the case for SC. It's 1 program so it's 1 mastermind which means that there's no way of choosing completely random numbers.
Thus in SC the "random" system is not random, there's a system behind it.
Top of the page   Quote  

2 months ago
#12

ToKa

   2
U can guess opponent location, when U see his color. Be4 when I made ~20 games per day i guess 80% on Python, but now with 10 games per week on other maps I cant.
Make 100+ games on one map and U will see U can guess on color.
It hapened to me too, 3 games 3 same positions.
Top of the page   Quote  

2 months ago
#13

g_G

   1
I agree start locations are not random, i get same locations so many times during consecutive games
Top of the page   Quote  

2 months ago
#14

Vaphell

   3
another thing - i played a lot with my friends RvR and i can't say I remember even single case of mirror match (chance of happening with true randomness 1/3 ?), i think all RvR games gave 2 different races. Something is wrong with SC randomness generator.
Top of the page   Quote  

2 months ago
#15

diniMuater

   9
2008-09-08 00:29:04, ETBrooD wrote:
#5 A computer has to decide which locations the players will get. So far so true, right? The conclusion would be that it cannot be random at all.
I'll explain to you what "random" really means: you have a set of cards that have numbers written on them, you turn them around so you can't see them, you look away from the cards, somebody else shuffles them, you turn back and then you choose one of the cards. That's random.
If you want this to work on a computer you must have at least 2 different programs running which isn't the case for SC. It's 1 program so it's 1 mastermind which means that there's no way of choosing completely random numbers.
Thus in SC the "random" system is not random, there's a system behind it.


hmmmm, not really. it remains the same, doesnt matter if its 1 process 2 processes or n processes. the rnadom generator needs a seed point, thats the actual random part of it. so you take a user action as seed for your random system, because thats an asynchonous event.

in school we had to make a die. it was only a 7 segment display which switched the numbers according to the clock and stoped after a few seconds. the outcome was compleatly random, coz the startpoint of this system was an asynchronous event from the user. very simple to make and very random.

for bw, the randomness is manipulated in some cases. the random race selection has some rules, if you search on google im sure you will find the exact way how it works. if i remember correct dt (crator of inhale) wrote a post about it on bwhacks.com.
EDIT: the thread

the startlocation randomness as you said is 1/1728 for your 3 games. there are soooo many games played so that 1/1728 isnt that seldom. but i never made any research on this, so i cant say anything for sure there.
Top of the page   Quote  

2 months ago
#16

fym.lpjuunin

   1
I've always thought something was "corrupt" with the random generator for scbw...

One good example is pro-gamer PvP's. Almost every PvP I've watched on youtube, like 85% of them have been brown vs orange. This pisses lots of people off, because in mirror matches, similar colors are a nuisance, on youtube you can barely tell the difference between the two, even with high quality on.

It gets to a point where when I see an interesting PvP vod I want to watch, before I click it, I silently say to myself "God, I bet you anything this is a freakin' brown vs orange..." I win the bet almost everytime...-.-(whoever heard of winning against god???) I also find myself complaining to my sc buddy alot, why is it that every PvP games in Korea are brown vs orange...we make fun of it sometimes claiming it's probably rigged to piss off tv/youtube viewers, but in all honesty, there has to be SOMETHING wrong here.

ETBrood and the OP is right, there can't be a random factor...

Another example is when I practice splitting on popular maps like python and bluestorm, sometimes I get same location over and over again (so I guess the theory of playing in same map consecutively fits).
Top of the page   Quote  

2 months ago
#17

ETBrooD

   2
#15 Your theory doesn't count for the SC engine. It's quite old so the random generator is old, too. At that time people didn't know so well how to program a tool that generates more random numbers. It will never be really random because the computer will always start counting from a certain point.
Of course technique has evolved so nowadays it's possible to get almost 100% randomness. Not 10+ years ago though.
Top of the page   Quote  

2 months ago
#18

Vaphell

   3
what? you speak as the ancient greeks wrote SC o_O

Ansi C - rand() function

"The ANSI C standard only states that rand() is a random number generator which generates integers in the range [0,RAND_MAX] inclusive, with RAND_MAX being a value defined in stdlib.h, and RAND_MAX being at least 32767. Note that 32767 isn't a very big number. If RAND_MAX is only 32767 then you can probably get only about 20,000 random numbers before the sequence starts to lose its randomness."
As you can see, standard C function gives you a lot of randomness and I assure you it existed 10 years ago.

Maybe computer generated sequences are not truly random, but not in a way you can get n times in a row the same result. Oldschool stuff like C64 or other 8bit junk could have 'random enough' generator so don't tell SC is too old to have good randomness, especially for a game. It's not as demanding as cryptography, some basic randomness will do.
It looks like the generator is seeded with not so unique seed and that's why combinations occur so often. Programmer's sloppiness or oversimplifying math formulas, not some technical barriers.
Top of the page   Quote  

2 months ago
#19

Economist_

   1
Agree with #18, especially with a seed, if you set a very simple seed for the generators then it may affect the randomness and if the seed is not generalized to include the extremely wide variation then you may come of with many times the results which have the same seed and therefore same outcome
Top of the page   Quote  

2 months ago
#20

Economist_

   1
Sorry for double post ^^
Top of the page   Quote  

1, 2 

Forum post
  • Do to others what you would have them do to you.
  • Do not judge so that you will not be judged.
    » Read posting rules in detail

    Reminder: IP-addresses for each comment is saved. Offensive posts will be censored. Repeated offense will result in a warning and/or a temporal lockdown of the account.
    • Are start locations truly random? \ Reply
    To submit a post, you have to be a registred member. Register an account below the title 'Membership' in the left menu.

    » Top of the page