I remember seeing this back in the HotS days as well... I think Random has a sort of weighting to it where it favors (but doesn't guarantee) the race you've played the least, and then when you've played them all an even number of times, you have a more even "die roll". Obviously Terran is your main race, but I would venture a guess you've played quite a bit of Zerg as well, which is why it's always picking Protoss.
I saw you can stack probes together, can you do that so when you scout the opponent at the start of the game, you can sneak one probe to the side, and the other probe just behave like a normal early game scout? So the one against you only see one probe, not two.
Stimmed marines without combat shield is dies in 2 adept hits instead of 3. In that voidray cheese game, in addition to not paying attention, stimming his marines lost him the fight.
My guess is that it is *truly* random, but people don't actually want that. Apple used to famously have real "random" play on iPods, but people thought it was buggy for playing the same band twice in a row, or the same song 3 times in an hour, so they made it LESS truly random and instead imposed a bunch of rules. SC2 players don't want true random, they want an even and regular spread of all three races. True random will clump up like this all the time.
The statistical chance of getting 3 or fewer Terran games out of 35 games if the chance of getting Terran in each game is truly random (33.3%) is only 0.068% or about 1 in 1500. So incredibly unlikely...
The thing with 1 in 1500 chances is that those do happen. Unlikely, but for example, if you play 3 random games/hr with 500 people you'd be extremely likely to have such a thing happen within a single day. Of course, specifically uThermal not getting Terran is incredibly -hilarious- _unlikely_ . 😂
31:45 idea: there is a rather large dark space on the main base. what if you dropped creep, drones, then made spine/spore? You could potentially march on the main nexus, then follow up with ling drops in the nat. Or something. With such a weird map I refuse to believe you can't figure something out!
Maybe random-ness is influenced by the total amount of people playing each race? For example, let’s say that following structure of race-choices is true: Terran - 27 Zerg - 18 Protos - 9 - In such case it is three times less probable for you to get terran than protoss and two times less probable to get terran over zerg. At least this idea came to my mind as a way of balancing the game / available choices by the AI.
If you mean put a 2nd gateway 'up' ie towards the rocks, then yeah it would block, but if the opp probe is clicked back home and takes a weird route then the opp will probably notice and realise the proxy was there still. The idea for this one though is to have 2 proxies so only 1 gets scouted and the opp thinks its only a 3gate (cos he saw only 2 at home).
It could also be that random just starts 3 queues (one for each races, with the mmr associated to each race) as selects the first best match found So it would be less "random" and more "no race requirement in search"
It would be cool to see you try compositions that are traditionally considered incompatible, like going hydra-corruptor vs toss air, add in lings +/- ultras, no banes
hello, my feeling in random back when i played sc2 (2011) was that you get a ponderation of the games played on your account, let say you play 80 terran 15% zerg 5% protoss, then your random algorithm will tend to compute probability so you converge to 33%,33%,33% is a reversed balaced manner ( so mostly P, then some Z, and almost never some T). And of course, in your case if this is a new account , i m completly and uterly wrong lol!
Ok hear me out: 1. Wheel of Fortune challenge -Two randomized wheels (preferably animated) determines one ground and one air unit. -You're only allowed to build those two units during matches, with exception to a small amount of starting game-units. 2. Centurion challenge -You may never have more than 100 supply. -That's it, good luck!
Getting the same race 3 times in a row is going to happen like 10% of sets-of-three games (1/3 you get it twice, and then 1/3 again to get it a third time, so 0.33*0.33 = 0.11), so it's not that implausible it would be rng.
Random doesn't mean what most people think it means. You can have random with prior knowledge, such as, drawing from a hat, and the hat refeshes only after it's empty. Or you can have random without prior knowledge, where each draw from the hat is immediately returned to the hat. People expect random to work like the first type, which will result in an equal distribution, but in true random, each draw is fresh, so it can be anything. In this case you could get an equal distribution, or all one race, or two races and none of the other. In fact, it would be unusual to get a completely equal distribution. Or flipping a coin: The chances of getting 10 heads in a row is the same as getting 5 heads followed by 5 tails.
Ha, Radhuset Station not being the best for a ling drop build vs Protoss is funny given the other day I watched Scarlett use a ling drop in the main and the pocket third paired with a nydus to completely overwhelm Astrea
@uThermal I've been playing random as well, and I get Protoss way more often than the other races, which is annoying. Would appreciate them to fix it to work better.
Chances of getting the same race 3 times in a row is always just 1/9, it's not that low when you make so many episodes, esp. comparing how low the odds are that they failed to call a random() function.
If it's completely random, you get a 1 in 3 chance each time. And probability to get same race for N times is 1/3 to power of N. This gives 1 in 9 chance to get Protoss three times in row. Unlikely, but can happen.
@@kaiser2261 , oops, yeap 1/27 of course. Still can happen. To make it *appear* more random blizz could change the formula to make it less likely to get same race as on previous play.
@@kaiser2261, actually you can't say that for sure without quite rigorous analysis. For example take a look at two Python functions below: import random import secrets def pairs_freq(s): freq = {} total = 0 for a, b in zip(s, s[1:]): k = a + b freq[k] = freq.get(k, 0) + 1 total += 1 for k, v in freq.items(): freq[k] = v / total return freq def show_freq(freq): for k in sorted(freq.keys()): print(f'{k}: {freq[k]:.2f}') First computes frequency for character pairs in a given string (e.g. for string "ababa" it would print "ab: 0.50 ba: 0.50"), second prints it out ordered. Let's run it on a string that encodes 50 games a friend of mine played as random: In [141]: show_freq(pairs_freq('zpttptzzzptttpzppttptzzzptttpzppttptzzzptttpzpttpt')) pp: 0.04 pt: 0.22 pz: 0.06 tp: 0.14 tt: 0.20 tz: 0.06 zp: 0.14 zz: 0.12 Looks pretty skewed right? Now, let's run it a few times with standard python random generator: In [133]: show_freq(pairs_freq([random.choice('tzp') for _ in range(0, 10)])) pz: 0.11 tp: 0.11 tt: 0.56 zt: 0.11 zz: 0.11 In [134]: show_freq(pairs_freq([random.choice('tzp') for _ in range(0, 10)])) pp: 0.22 pz: 0.11 tt: 0.33 tz: 0.11 zt: 0.22 For small sequences you get pretty skewed results as well. In fact, even for a string of 100 characters it is quite imbalanced: show_freq(pairs_freq([random.choice('tzp') for _ in range(0, 100)])) pp: 0.09 pt: 0.11 pz: 0.07 tp: 0.10 tt: 0.14 tz: 0.15 zp: 0.08 zt: 0.14 zz: 0.11 Results start to be uniform for N closer to 10,000. Actually, even if cryptographic random number generator from secrets package is used it looks pretty skewed on 100: In [139]: show_freq(pairs_freq([secrets.choice('tzp') for _ in range(0, 100)])) pp: 0.06 pt: 0.15 pz: 0.12 tp: 0.15 tt: 0.08 tz: 0.11 zp: 0.12 zt: 0.11 zz: 0.09 => random as percepted by human is not the same as truly random :)
video idea: - live coach a silver or gold player and see how big of a win streak you can get. Id love to be told what to do with my 70 apm and the hilariousness that would follow :p
Bro as a random player I always get mirror match ups or Protoss 75 percent of the time. What about an option to play random but no mirror match ups option and a just mirror match up option. Would be great for the game IMO
Your in Mexico, ask your gf since she is from Mexico about it hot honey, lime and tequila shot, helps a lot with sickness trust me. It’s the Mexican cure for the cold!!!
This is how random is supose to work marc. Youre gonna get one race more than others. The chance lowers the more it happens in a row but still possible when there are only 3 options.
I wonder if the random algorithm is trying to balance out the server. Like if not many people are playing protoss that day it gives you protoss 90% of the time or something.
most stuff we call "random" is 1 - 3 of same interval, Like this: p,p,t,z,t,p,z,z,z,t,p,t,z,p,z,t,p true random can be real ugly, Like this: p,z,t,p,p,p,p,p,p,p,z,z,t,t,p,t,t,t,z,t,t,p,t,t,z (this rare, but probable)
Man this is how all of my PvP have alwasy been. I always 1 base all-in vs other protoss and don't see the point in ever expanding and have over 99% win rate in protoss mirror. Yeah, that's a sick DT rush strat and still gets the GG at 5:20. Sick.
Dear thermos, after seeing you compete I think you need to cast a good few high level matches. The meta at the top seems to mix greed and minimum defense/damage in new ways not seen before. You are wonderful at doing your own thing but for almost everyone else sc2 is a game of studying other players tricks n treats.
Marc stealing his opponent's gas: "I love being annoying."
The opponent stealing Marc's gas: _surprised Pikachu face_
Would love a timestamp for that 😂
@@pokerstar2926 7:44 There you go, and enjoy! ^-^
Based on an analysis of the last 2,000 games, I found that Marc's worst match-up is Protoss vs Mosquito.
5:57 Mosquito: "GG Marc. May I have some blood?"
7:29 Mosquito: "How about now?"
23:11 Mosquito: "Please?"
Now imagine if that mosquito had malaria.
In mother russia you drink the mosquitoe blood
I love how we've gotten to the point where he doesn't even announce when he puts the Nexus in his opponent's base 😂😂😂
Right? Even watching the video i totally missed that. How was the opponent supposed to know about the nexus if even the viewer didn't lol
On this map, scarlett did 8 ling drop in main then nydus with queens and followed up with swarm host. It was wild.
I remember seeing this back in the HotS days as well... I think Random has a sort of weighting to it where it favors (but doesn't guarantee) the race you've played the least, and then when you've played them all an even number of times, you have a more even "die roll". Obviously Terran is your main race, but I would venture a guess you've played quite a bit of Zerg as well, which is why it's always picking Protoss.
I saw you can stack probes together, can you do that so when you scout the opponent at the start of the game, you can sneak one probe to the side, and the other probe just behave like a normal early game scout? So the one against you only see one probe, not two.
Ooooooooh!!
Sewer mermaid strats
Yep you can, and yep he and other people has, and works quite well even against observant players :D
“I’m going to steal his gas because it’s annoying and I like to be annoying.” 😂😂😂
ive not played SC2 in years but i do enjoy watched video's like this. keep it up!
Stimmed marines without combat shield is dies in 2 adept hits instead of 3. In that voidray cheese game, in addition to not paying attention, stimming his marines lost him the fight.
My guess is that it is *truly* random, but people don't actually want that. Apple used to famously have real "random" play on iPods, but people thought it was buggy for playing the same band twice in a row, or the same song 3 times in an hour, so they made it LESS truly random and instead imposed a bunch of rules. SC2 players don't want true random, they want an even and regular spread of all three races. True random will clump up like this all the time.
For example, if you flip a coin 100 times in a row, getting 6 or even more heads or tails in a row is not at all uncommon.
Awe, I hope you feel better Thermy!!! ♥️♥️♥️
That second game was just stupendous
The statistical chance of getting 3 or fewer Terran games out of 35 games if the chance of getting Terran in each game is truly random (33.3%) is only 0.068% or about 1 in 1500. So incredibly unlikely...
Isn't that chance three times higher 'cos it could happen with different races also? It would get same attention if he would get only zerg etc.
The thing with 1 in 1500 chances is that those do happen. Unlikely, but for example, if you play 3 random games/hr with 500 people you'd be extremely likely to have such a thing happen within a single day.
Of course, specifically uThermal not getting Terran is incredibly -hilarious- _unlikely_ . 😂
What are you asking?
@Salmacream
Y r u gae?
Isn't that 1/431439 chance?
31:45 idea: there is a rather large dark space on the main base. what if you dropped creep, drones, then made spine/spore? You could potentially march on the main nexus, then follow up with ling drops in the nat. Or something. With such a weird map I refuse to believe you can't figure something out!
Really enjoying this delicious Protoss cheese platter that you kindly served up!
This Nexus in opponent base is a famous strategy that SoS once did successfully. Exactly the same as Thermy did in this video.
I have also experienced the "random is not random" effect. Much higher probability of getting the same again than getting another one.
Wow that second game made me laugh my ass off. 😂 You are a psychopath to go for a cheese like that and I cannot believe how effective it was.
International flights seem guaranteed to make people sick. Feel better soon!
ForGG's content was recommended to me yesterday, that guy is super chill. Not what I expected.
Does he still make new content? I miss his hellion builds. Sad nostalgia vibes
Maybe random-ness is influenced by the total amount of people playing each race?
For example, let’s say that following structure of race-choices is true:
Terran - 27
Zerg - 18
Protos - 9
- In such case it is three times less probable for you to get terran than protoss and two times less probable to get terran over zerg. At least this idea came to my mind as a way of balancing the game / available choices by the AI.
3:22 if you close off that path with a 2nd gateway, would the opp probe take a different route and miss those 2 proxies?
If you mean put a 2nd gateway 'up' ie towards the rocks, then yeah it would block, but if the opp probe is clicked back home and takes a weird route then the opp will probably notice and realise the proxy was there still. The idea for this one though is to have 2 proxies so only 1 gets scouted and the opp thinks its only a 3gate (cos he saw only 2 at home).
It could also be that random just starts 3 queues (one for each races, with the mmr associated to each race) as selects the first best match found
So it would be less "random" and more "no race requirement in search"
No, random MMR is not race specific.
It would be cool to see you try compositions that are traditionally considered incompatible, like going hydra-corruptor vs toss air, add in lings +/- ultras, no banes
any chance for analysis of Clem's sensational TvZ in DH:Atlanta? he played out of his mind
Wow, game 2 with the SoS callback! Didn't think that strat was possible more than one time
hello, my feeling in random back when i played sc2 (2011) was that you get a ponderation of the games played on your account, let say you play 80 terran 15% zerg 5% protoss, then your random algorithm will tend to compute probability so you converge to 33%,33%,33% is a reversed balaced manner ( so mostly P, then some Z, and almost never some T).
And of course, in your case if this is a new account , i m completly and uterly wrong lol!
International flight = Contamination room 🤣🤣🤣
I am so sorry you are not feeling good, take good care of your self!
I like cannon rush + proxy robo against terran, just need to get immortals ready when they get tanks, maybe feenix would also work well against tanks
Ok hear me out:
1. Wheel of Fortune challenge
-Two randomized wheels (preferably animated) determines one ground and one air unit.
-You're only allowed to build those two units during matches, with exception to a small amount of starting game-units.
2. Centurion challenge
-You may never have more than 100 supply.
-That's it, good luck!
I hope you get well soon
Oo ee oo ah ah ting tang walla walla bing bang
Feel better Marc!
at 6:40, you were looking for the word "occasional"
@uThermal what if on the very large map that isn't good for zerg you just take the extra time and fly the overlords into the main base?
Thanks for the fun. ^^
That Serral v Reynor game was disgusting... Reynor played a ridiculous game.
Getting the same race 3 times in a row is going to happen like 10% of sets-of-three games (1/3 you get it twice, and then 1/3 again to get it a third time, so 0.33*0.33 = 0.11), so it's not that implausible it would be rng.
I am sure the terran broke the F10 N keys and left the game
Random doesn't mean what most people think it means. You can have random with prior knowledge, such as, drawing from a hat, and the hat refeshes only after it's empty. Or you can have random without prior knowledge, where each draw from the hat is immediately returned to the hat. People expect random to work like the first type, which will result in an equal distribution, but in true random, each draw is fresh, so it can be anything. In this case you could get an equal distribution, or all one race, or two races and none of the other. In fact, it would be unusual to get a completely equal distribution. Or flipping a coin: The chances of getting 10 heads in a row is the same as getting 5 heads followed by 5 tails.
Think that was Jason, saw this on stream lol he was so mad at himself
Yep that's definitely Jason lol
Can you please give us a link
@@hayden1342www.twitch.tv/videos/2005163302 at about 30 mins, it's quite funny.
@@hayden1342 www.twitch.tv/videos/2005163302 30 minutes in
@@hayden1342 And if the link gets spam blocked then look up JuggernautJason
on twitch and see the VOD from 16th december 30 minutes in.
Ha, Radhuset Station not being the best for a ling drop build vs Protoss is funny given the other day I watched Scarlett use a ling drop in the main and the pocket third paired with a nydus to completely overwhelm Astrea
I love your enthusiasm. I cannot place your accent? Finnish?
@uThermal I've been playing random as well, and I get Protoss way more often than the other races, which is annoying. Would appreciate them to fix it to work better.
The SOS Deja Vu
Didn't you play nydus swarm hosts against protoss on Radhuset Station in one of the earlier videos and it was really strong? Maybe it was terran..
Forgot to wash down the ceviche with sterilizing tequila 😂 get well soon 😊
those were some SICK cheeses (pun intended)
Does someone have a list of the races he got and in order? We can do a run rate check on the statistical validity of them
Chances of getting the same race 3 times in a row is always just 1/9, it's not that low when you make so many episodes, esp. comparing how low the odds are that they failed to call a random() function.
Marc, my comment about the "bug" and "only zerg" was just a joke, I don't have _any_ knowledge of a bug, I made that completely up.
If it's completely random, you get a 1 in 3 chance each time. And probability to get same race for N times is 1/3 to power of N. This gives 1 in 9 chance to get Protoss three times in row. Unlikely, but can happen.
1 in 27. And its an issue being reported by multiple players.
@@kaiser2261 , oops, yeap 1/27 of course. Still can happen. To make it *appear* more random blizz could change the formula to make it less likely to get same race as on previous play.
@@eddyz6502 its a bug that needs to be fixed its not the appearance thats off
@@kaiser2261, actually you can't say that for sure without quite rigorous analysis.
For example take a look at two Python functions below:
import random
import secrets
def pairs_freq(s):
freq = {}
total = 0
for a, b in zip(s, s[1:]):
k = a + b
freq[k] = freq.get(k, 0) + 1
total += 1
for k, v in freq.items():
freq[k] = v / total
return freq
def show_freq(freq):
for k in sorted(freq.keys()):
print(f'{k}: {freq[k]:.2f}')
First computes frequency for character pairs in a given string (e.g. for string "ababa" it would print "ab: 0.50 ba: 0.50"), second prints it out ordered.
Let's run it on a string that encodes 50 games a friend of mine played as random:
In [141]: show_freq(pairs_freq('zpttptzzzptttpzppttptzzzptttpzppttptzzzptttpzpttpt'))
pp: 0.04
pt: 0.22
pz: 0.06
tp: 0.14
tt: 0.20
tz: 0.06
zp: 0.14
zz: 0.12
Looks pretty skewed right?
Now, let's run it a few times with standard python random generator:
In [133]: show_freq(pairs_freq([random.choice('tzp') for _ in range(0, 10)]))
pz: 0.11
tp: 0.11
tt: 0.56
zt: 0.11
zz: 0.11
In [134]: show_freq(pairs_freq([random.choice('tzp') for _ in range(0, 10)]))
pp: 0.22
pz: 0.11
tt: 0.33
tz: 0.11
zt: 0.22
For small sequences you get pretty skewed results as well.
In fact, even for a string of 100 characters it is quite imbalanced:
show_freq(pairs_freq([random.choice('tzp') for _ in range(0, 100)]))
pp: 0.09
pt: 0.11
pz: 0.07
tp: 0.10
tt: 0.14
tz: 0.15
zp: 0.08
zt: 0.14
zz: 0.11
Results start to be uniform for N closer to 10,000.
Actually, even if cryptographic random number generator from secrets package is used it looks pretty skewed on 100:
In [139]: show_freq(pairs_freq([secrets.choice('tzp') for _ in range(0, 100)]))
pp: 0.06
pt: 0.15
pz: 0.12
tp: 0.15
tt: 0.08
tz: 0.11
zp: 0.12
zt: 0.11
zz: 0.09
=> random as percepted by human is not the same as truly random :)
maybe it's using the windows random function :P
that killed millions in crypto
OMG this is so entertaining, I have to stop watching because I need to work, hahaha! Will come back later, lol
needs more kippensoep
Random always rolled the same race for me 5x in a row.
I only play random, I get all 3 races often. I actually get Terran and Zerg a bit more than Protoss
Cool R̶a̶n̶d̶o̶m̶ Protoss cheese to grandmaster!
video idea: - live coach a silver or gold player and see how big of a win streak you can get. Id love to be told what to do with my 70 apm and the hilariousness that would follow :p
Bro as a random player I always get mirror match ups or Protoss 75 percent of the time. What about an option to play random but no mirror match ups option and a just mirror match up option. Would be great for the game IMO
Your in Mexico, ask your gf since she is from Mexico about it hot honey, lime and tequila shot, helps a lot with sickness trust me. It’s the Mexican cure for the cold!!!
This is how random is supose to work marc. Youre gonna get one race more than others. The chance lowers the more it happens in a row but still possible when there are only 3 options.
the chance never changes--1/3 is 1/3 is 1/3.
@theconiferoust9598 Yeah I meant probability. The chance remains the same, aye.
I played random 1v1 the other day and got Terran 7 times in a row lol
I wonder if the random algorithm is trying to balance out the server. Like if not many people are playing protoss that day it gives you protoss 90% of the time or something.
I challenge you to make a video i watch where I fail to learn CRITICAL information on survival.
uthermal, do paneling burrowed mines!
sending oracles to "revelate" 😅
There are no bugs. Protoss is the most random race.
f every cheese. As terran I check the fog aside main base every match and put my third depot there, cuz I was so f screwed by those cheesers.
bANGIN' !
Honestly I'd be cool if the title included the name of the series, at least let us known in just the thumbnail. Love random to rank 1, cheers
most stuff we call "random" is 1 - 3 of same interval,
Like this: p,p,t,z,t,p,z,z,z,t,p,t,z,p,z,t,p
true random can be real ugly,
Like this: p,z,t,p,p,p,p,p,p,p,z,z,t,t,p,t,t,t,z,t,t,p,t,t,z (this rare, but probable)
Wohooooo
Can you stream on UA-cam too. Twitch let creators co stream a few months ago
Dude this is so true about random, I used to play it and we called it Rantoss because I were getting like 80% P. I glad I am not only one
I miss you Marc when are we getting a new video?
Let's go!
Игра в снежки, ха ха :)
That first game against a random ass Protoss was a really fast win.
no such thing as for or disgusx or bad or etc, outx any nmw s perfx
Random just not giving you Terran and mostly Protoss is kinda ruining the challenge lol
sir, every new video is the most x thing ever. you're forcing me to take the position of grumpy old man
That’s just the nature of UA-cam now, every title is click bait, every video image is of the creator looking surprised or a big explosion, etc.
I swear, how have you gotten Protoss almost every time? I've watched like 5 videos and you got terran once and zerg twice
No videos anymore? 😢
DELICIOUS ❤
Man this is how all of my PvP have alwasy been. I always 1 base all-in vs other protoss and don't see the point in ever expanding and have over 99% win rate in protoss mirror.
Yeah, that's a sick DT rush strat and still gets the GG at 5:20. Sick.
Anychance you could let us know which part of mexico you are in?
I play random and all I get is protoss on na server here 4500 mmr
Is he on vacation?
Siege warfare to gm pls
U can be anoing while containing 🎉
Blizzard does not exist anymore. That's why there is silly bugs in 13-years old game.
Dear thermos, after seeing you compete I think you need to cast a good few high level matches. The meta at the top seems to mix greed and minimum defense/damage in new ways not seen before. You are wonderful at doing your own thing but for almost everyone else sc2 is a game of studying other players tricks n treats.
you're free to watch literally any other sc2 channel. plenty of high level players casting meta shit, only one thermy.
@@nonstandard5492 That wouldn't be the same.
❤
2nd
Первый!
Omg so much tasty cheesseeeee, provolone 🫦, mascarpone, parmesan, can cheesse. Me gusta
intermittent @uthermal
Let's go!