• Log InLog In
  • Register
Liquid`
Team Liquid Liquipedia
EDT 07:41
CEST 13:41
KST 20:41
  • Home
  • Forum
  • Calendar
  • Streams
  • Liquipedia
  • Features
  • Store
  • EPT
  • TL+
  • StarCraft 2
  • Brood War
  • Smash
  • Heroes
  • Counter-Strike
  • Overwatch
  • Liquibet
  • Fantasy StarCraft
  • TLPD
  • StarCraft 2
  • Brood War
  • Blogs
Forum Sidebar
Events/Features
News
Featured News
Team Liquid Map Contest #22 - The Finalists9[ASL21] Ro16 Preview Pt1: Fresh Flow9[ASL21] Ro24 Preview Pt2: News Flash10[ASL21] Ro24 Preview Pt1: New Chaos0Team Liquid Map Contest #22 - Presented by Monster Energy21
Community News
2026 GSL Season 1 Qualifiers6Maestros of the Game 2 announced22026 GSL Tour plans announced4Weekly Cups (April 6-12): herO doubles, "Villains" prevail0MaNa leaves Team Liquid18
StarCraft 2
General
Team Liquid Map Contest #22 - The Finalists 2026 GSL Tour plans announced Maestros of the Game 2 announced Weekly Cups (April 6-12): herO doubles, "Villains" prevail MaNa leaves Team Liquid
Tourneys
Sparkling Tuna Cup - Weekly Open Tournament 2026 GSL Season 1 Qualifiers Master Swan Open (Global Bronze-Master 2) SEL Doubles (SC Evo Bimonthly) $5,000 WardiTV TLMC tournament - Presented by Monster Energy
Strategy
Custom Maps
[D]RTS in all its shapes and glory <3 [A] Nemrods 1/4 players [M] (2) Frigid Storage
External Content
Mutation # 521 Memorable Boss The PondCast: SC2 News & Results Mutation # 520 Moving Fees Mutation # 519 Inner Power
Brood War
General
ASL21 General Discussion BGH Auto Balance -> http://bghmmr.eu/ Data needed A cwal.gg Extension - Easily keep track of anyone [ASL21] Ro16 Preview Pt1: Fresh Flow
Tourneys
[ASL21] Ro16 Group B [ASL21] Ro16 Group A [ASL21] Ro24 Group F [Megathread] Daily Proleagues
Strategy
What's the deal with APM & what's its true value Any training maps people recommend? Fighting Spirit mining rates Muta micro map competition
Other Games
General Games
Battle Aces/David Kim RTS Megathread Nintendo Switch Thread Stormgate/Frost Giant Megathread General RTS Discussion Thread Starcraft Tabletop Miniature Game
Dota 2
The Story of Wings Gaming Official 'what is Dota anymore' discussion
League of Legends
G2 just beat GenG in First stand
Heroes of the Storm
Simple Questions, Simple Answers Heroes of the Storm 2.0
Hearthstone
Deck construction bug Heroes of StarCraft mini-set
TL Mafia
Vanilla Mini Mafia Mafia Game Mode Feedback/Ideas TL Mafia Community Thread Five o'clock TL Mafia
Community
General
Things Aren’t Peaceful in Palestine US Politics Mega-thread Canadian Politics Mega-thread Russo-Ukrainian War Thread European Politico-economics QA Mega-thread
Fan Clubs
The IdrA Fan Club
Media & Entertainment
[Req][Books] Good Fantasy/SciFi books [Manga] One Piece Movie Discussion!
Sports
2024 - 2026 Football Thread Formula 1 Discussion Cricket [SPORT] Tokyo Olympics 2021 Thread
World Cup 2022
Tech Support
[G] How to Block Livestream Ads
TL Community
The Automated Ban List
Blogs
Reappraising The Situation T…
TrAiDoS
lurker extra damage testi…
StaticNine
Broowar part 2
qwaykee
Funny Nicknames
LUCKY_NOOB
Iranian anarchists: organize…
XenOsky
ASL S21 English Commentary…
namkraft
Customize Sidebar...

Website Feedback

Closed Threads



Active: 2307 users

David Kim's Current Balance Thoughts - Page 58

Forum Index > SC2 General
1229 CommentsPost a Reply
Prev 1 56 57 58 59 60 62 Next
samurai80
Profile Joined November 2011
Japan4225 Posts
January 23 2014 11:08 GMT
#1141
On January 23 2014 18:35 Chaggi wrote:
Show nested quote +
On January 23 2014 18:29 samurai80 wrote:
On January 23 2014 18:26 Chaggi wrote:
On January 23 2014 18:22 samurai80 wrote:
On January 23 2014 18:12 Chaggi wrote:
On January 23 2014 18:10 samurai80 wrote:
On January 23 2014 18:07 Chaggi wrote:
On January 23 2014 18:06 samurai80 wrote:
On January 23 2014 17:54 plogamer wrote:
On January 23 2014 17:48 samurai80 wrote:
[quote]
This is precisely what I meant in my previous post.

Where were you when they nerfed hellbats and widowmines?

But they did buff mech/air upgrades.


which has done absolutely nothing for the MU that people care about, TvP. Yay!

Mech is becoming a thing. But, I was actually against the hellbat nerf. I liked PvT when terrans had the allmighty hellbats much more to be fair (even if they looked op sometimes ).


Mech is never going to become a "thing" in TvP unless some massive changes happen.

Hence what I said about hellbats.


I'm not sure a unnerfed hellbat would be the massive changes that need to happen.

The thing is the m-u actually WAS better before the nerf, which led me to the conclusion that it should be unnerfed for TvP (and the fix was aimed at TvT).


I'm not sure it was simply just that. It was partially better cause Protoss was still struggling to defend stuff like boosted medivacs, and you had interesting compositions like marauder/hellbat that ForGG was known for.

Well maybe you're right, tbh I dunno how things should be fixed.
z0rz
Profile Blog Joined June 2011
United States350 Posts
Last Edited: 2014-01-23 11:22:48
January 23 2014 11:22 GMT
#1142
On January 23 2014 16:08 GolemMadness wrote:
Let's all go crazy over a small sample size of games from the GSL! Meanwhile, terran has a 64% win rate in Proleague, while protoss has a win rate of 44%.

Not sure if serious. Of the 55 players to participate in Proleague so far, only 12 are Terran. And only 3 of those 12 are actually winning a significant amount of games.
twitch.tv/fartymcbutt
hfsrj
Profile Joined October 2012
Germany166 Posts
January 23 2014 11:39 GMT
#1143
I know it's rude to auto-quote, but I'd like to have some comments on this :

On January 23 2014 04:55 hfsrj wrote:
Btw I don't know if they noticed that most way to "correct for skill" actually hide the imbalance, especially if the most powerfull race is more represented than the two other.

It's quite easy to show : let's say it's 25%A, 25%B, 50%C, for the sake of the argument. If C is favored, and you are A, you loose most of your games versus C, for equal skill (say 100% for simplicity, the argument is still valid for lower). Then you proceed to loose 50% of your games against A and B (same skill). You therefore lost 75% of your games, and your MMR (perceived skill) goes down untill you reach 50% win ratio, i.e. your perceived skill is lower than actual skill.

Things are more complicated in truth, but the statistics still apply. If they count skill by MMR, the figure they show are erroneous. This phenomena exist if C is not overrepresented, but it is obviously smaller, and much harder to correct for.


Anyone knows how the skill correction is made ? It's a bit hard not to be biased in doing it, so I'd really like to know more. I'm a physicist and quite comfortable with statistics, btw, and I think it'd be nice in this topic to have more arguments than "small sample size" and "corrected win percentage", that are actually not very informative.
Daswollvieh
Profile Blog Joined October 2009
5553 Posts
January 23 2014 12:00 GMT
#1144
And now, make it four ST players in Code S.
Big J
Profile Joined March 2011
Austria16289 Posts
January 23 2014 12:01 GMT
#1145
On January 23 2014 21:00 Daswollvieh wrote:
And now, make it four ST players in Code S.


ahm... still a game to be played...
Bagi
Profile Joined August 2010
Germany6799 Posts
January 23 2014 12:03 GMT
#1146
On January 23 2014 21:00 Daswollvieh wrote:
And now, make it four ST players in Code S.

He still has to overcome the biggest obstacle of them all: a protoss.
Plansix
Profile Blog Joined April 2011
United States60190 Posts
January 23 2014 12:05 GMT
#1147
On January 23 2014 21:03 Bagi wrote:
Show nested quote +
On January 23 2014 21:00 Daswollvieh wrote:
And now, make it four ST players in Code S.

He still has to overcome the biggest obstacle of them all: a protoss.

You heard it hear folks, zerg under powered vs terran. Blizzard please nerf.
I have the Honor to be your Obedient Servant, P.6
TL+ Member
vthree
Profile Joined November 2011
Hong Kong8039 Posts
January 23 2014 12:06 GMT
#1148
lol, it is so funny to see Protoss players try to justify TvP balance by saying all the Terrans that haven't played KR protosses are actually TvP bonjwas. You know, MMA, ForGG, MKP, Ensnare , Virus, TOP, Goody, Thorzain, IMHappy, Ganzi etc... If only they were playing in GSL, we would have balance race ratios. They are all much better than Innovation!!!

Protoss were actually trying to build Hack, SuperNova up as tier 1 terrans...
Destructicon
Profile Blog Joined September 2011
4713 Posts
January 23 2014 12:07 GMT
#1149
On January 23 2014 21:00 Daswollvieh wrote:
And now, make it four ST players in Code S.


Startale refusing to die even though they have lost their most iconic player and coaching staff, fighting spirit right there. Come on Hack, you can do it!
WriterNever give up, never surrender! https://www.youtube.com/user/DestructiconSC
vthree
Profile Joined November 2011
Hong Kong8039 Posts
Last Edited: 2014-01-23 12:12:08
January 23 2014 12:09 GMT
#1150
On January 23 2014 18:26 Chaggi wrote:
Show nested quote +
On January 23 2014 18:22 samurai80 wrote:
On January 23 2014 18:12 Chaggi wrote:
On January 23 2014 18:10 samurai80 wrote:
On January 23 2014 18:07 Chaggi wrote:
On January 23 2014 18:06 samurai80 wrote:
On January 23 2014 17:54 plogamer wrote:
On January 23 2014 17:48 samurai80 wrote:
On January 23 2014 17:46 ejozl wrote:
IMO balance today is at it's worst it ever was. At least during BL inf times timings could be hit to deny that zerg deathball. Bit by bit every viable timing of terran has been nerfed out of the game, so that protoss and zerg dominate anything past 15 minutes.

1 month of Terran not knowing what to do in PvT does not equal the Broodlordwar Era.
RTS is almost self-balancing in the way that players who do worse, because of their race or certain "imbalances," figure out something new, in order to win. I don't like the approach of buffing units like the Ghost to tell players how they should play.

This is precisely what I meant in my previous post.

Where were you when they nerfed hellbats and widowmines?

But they did buff mech/air upgrades.


which has done absolutely nothing for the MU that people care about, TvP. Yay!

Mech is becoming a thing. But, I was actually against the hellbat nerf. I liked PvT when terrans had the allmighty hellbats much more to be fair (even if they looked op sometimes ).


Mech is never going to become a "thing" in TvP unless some massive changes happen.

Hence what I said about hellbats.


I'm not sure a unnerfed hellbat would be the massive changes that need to happen.


Actually, unnerfed hellbat would be a huge deal... Because PO actually doesn't hurt hellbats much... Protoss would actually need to build up an army before going AoE. This game is always about timing. There is no way for Protoss to get an easy third with minimal army if the old hellbats were here.

RTS be nature are snowball-y... Just having a 60 sec window where terran can do damage is a HUGE deal... That is why the Queen change made BL/infestors so strong. Pros are so good that even a 30 second advantage can be enough for them to snowball to a win.
vthree
Profile Joined November 2011
Hong Kong8039 Posts
January 23 2014 12:13 GMT
#1151
On January 23 2014 20:22 z0rz wrote:
Show nested quote +
On January 23 2014 16:08 GolemMadness wrote:
Let's all go crazy over a small sample size of games from the GSL! Meanwhile, terran has a 64% win rate in Proleague, while protoss has a win rate of 44%.

Not sure if serious. Of the 55 players to participate in Proleague so far, only 12 are Terran. And only 3 of those 12 are actually winning a significant amount of games.


Because Proleague coaches don't want to win and Protoss are better in Bo1? Except they are dominating Bo3s as well?
Phobos_
Profile Joined January 2014
Canada1 Post
Last Edited: 2014-01-23 12:22:53
January 23 2014 12:16 GMT
#1152
On January 23 2014 20:39 hfsrj wrote:
I know it's rude to auto-quote, but I'd like to have some comments on this :

Show nested quote +
On January 23 2014 04:55 hfsrj wrote:
Btw I don't know if they noticed that most way to "correct for skill" actually hide the imbalance, especially if the most powerfull race is more represented than the two other.

It's quite easy to show : let's say it's 25%A, 25%B, 50%C, for the sake of the argument. If C is favored, and you are A, you loose most of your games versus C, for equal skill (say 100% for simplicity, the argument is still valid for lower). Then you proceed to loose 50% of your games against A and B (same skill). You therefore lost 75% of your games, and your MMR (perceived skill) goes down untill you reach 50% win ratio, i.e. your perceived skill is lower than actual skill.

Things are more complicated in truth, but the statistics still apply. If they count skill by MMR, the figure they show are erroneous. This phenomena exist if C is not overrepresented, but it is obviously smaller, and much harder to correct for.


Anyone knows how the skill correction is made ? It's a bit hard not to be biased in doing it, so I'd really like to know more. I'm a physicist and quite comfortable with statistics, btw, and I think it'd be nice in this topic to have more arguments than "small sample size" and "corrected win percentage", that are actually not very informative.


The only way (that we know of anyway) that Blizzard can measure skill is the mmr system. So most likely what they are doing is similar to how they update the mmr of players, some kind of Bayesian update. When two players play a game the result gives us more information about the skill or mmr of each player. Bayes' rule tells us how to compute this. It is kind of like a weighting, if player A is much more skilled then player B and player A wins then the system does not change very much because the outcome was expected with high probability. However, if player B wins then that is not expected and a larger change to the players mmrs would occur.

You can do a similar thing with race match ups. If a masters Zerg plays a masters Terran then the system knows the skills of each player and so it has some kind of idea on what the expected outcome is. So if the Zerg player has a much higher mmr than the Terran and the Zerg wins then the system would not consider that very good evidence that Zerg has a higher win rate than Terran at a similar skill level in the masters league, ie skill is getting factored out. You can actually make this mathematically rigorous pretty easily.

The problem with this is that imbalance in the game gets incorporated into a player's skill or mmr after they play some games. If you give marines like 10 range, after all the Terran's play a bunch of games their mmr or "skill" is going to go up because they will start beating better players. We as people know this is imbalance but the system doesn't differentiate because it can't know the difference. So basically when they say they are factoring out skill they are most likely factoring out imbalance at the same time.

If this is actually what they are doing I find it pretty comical that they use this along with Proleague results to justify anything. The results from Proleague are basically just as statistically ridiculous because there are so few Terrans in Proleague and a huge number of the Terran wins come from only a couple players so it says basically nothing about TvP statistically.
vthree
Profile Joined November 2011
Hong Kong8039 Posts
January 23 2014 12:19 GMT
#1153
On January 23 2014 17:46 ejozl wrote:
Show nested quote +
IMO balance today is at it's worst it ever was. At least during BL inf times timings could be hit to deny that zerg deathball. Bit by bit every viable timing of terran has been nerfed out of the game, so that protoss and zerg dominate anything past 15 minutes.

1 month of Terran not knowing what to do in PvT does not equal the Broodlordwar Era.
RTS is almost self-balancing in the way that players who do worse, because of their race or certain "imbalances," figure out something new, in order to win. I don't like the approach of buffing units like the Ghost to tell players how they should play.



1 month??? WTF are you smoking?

Fine, let's play all code S PvZ games be played on Daedulus and see how the protoss pros self-balance?

Do you think WoL was self balanced? WoL Thor nerf, ghost nerf, bunker nerf, BFH nerf. HoTs, mine nerf, hellbats nerf...

No buffs? Why were oracles buffs? Why were immortals buff? Protoss upgrade buffed?
Slydie
Profile Joined August 2013
1935 Posts
January 23 2014 12:24 GMT
#1154
I have seen the thing!

LotV is a toss based expansion, and what is more natural than having P overpowered before the release? Maybe there will be som nerfs right before the release like with Zerg before HotS.

Also, if we are all happy, why should we ever buy LotV? That is Blizzards main concern now. However, there is a big risk, LotV better be absolutely amazing and put SC2 into the history books with amazing game design and balance, the way BW did...
Buff the siegetank
Destructicon
Profile Blog Joined September 2011
4713 Posts
January 23 2014 12:24 GMT
#1155
On January 23 2014 21:16 Phobos_ wrote:
Show nested quote +
On January 23 2014 20:39 hfsrj wrote:
I know it's rude to auto-quote, but I'd like to have some comments on this :

On January 23 2014 04:55 hfsrj wrote:
Btw I don't know if they noticed that most way to "correct for skill" actually hide the imbalance, especially if the most powerfull race is more represented than the two other.

It's quite easy to show : let's say it's 25%A, 25%B, 50%C, for the sake of the argument. If C is favored, and you are A, you loose most of your games versus C, for equal skill (say 100% for simplicity, the argument is still valid for lower). Then you proceed to loose 50% of your games against A and B (same skill). You therefore lost 75% of your games, and your MMR (perceived skill) goes down untill you reach 50% win ratio, i.e. your perceived skill is lower than actual skill.

Things are more complicated in truth, but the statistics still apply. If they count skill by MMR, the figure they show are erroneous. This phenomena exist if C is not overrepresented, but it is obviously smaller, and much harder to correct for.


Anyone knows how the skill correction is made ? It's a bit hard not to be biased in doing it, so I'd really like to know more. I'm a physicist and quite comfortable with statistics, btw, and I think it'd be nice in this topic to have more arguments than "small sample size" and "corrected win percentage", that are actually not very informative.


The only way (that we know of anyway) that Blizzard can measure skill is the mmr system. So most likely what they are doing is similar to how they update the mmr of players, some kind of Bayesian update. When two players play a game the result gives us more information about the skill or mmr of each player. Bayes' rule tells us how to compute this. It is kind of like a weighting, if player A is much more skilled then player B and player A wins then the system does not change very much because the outcome was expected with high probability. However, if player B wins then that is not expected and a larger change to the players mmrs would occur.

You can do a similar thing with race match ups. If a masters Zerg plays a masters Terran then the system knows the skills of each player and so it has some kind of idea on what the expected outcome is. So if the Zerg player has a much higher mmr than the Terran and the zerg wins then the system would not consider that very good evidence that Zerg has a higher win rate than Terran at a similar skill level in the masters league, ie skill is getting factored out. You can actually make this mathematically rigorous pretty easily.

The problem with this is that imbalance in the game gets incorporated into a player's skill or mmr after they play some games. If you give marines like 10 range, after all the Terran's play a bunch of games their mmr or "skill" is going to go up because they will start beating better players. We as people know this is imbalance but the system doesn't differentiate because it can't know the difference. So basically when they say they are factoring out skill they are most likely factoring out imbalance at the same time.

If this is actually what they are doing I find it pretty comical that they use this along with Proleague results to justify anything. The results from Proleague are basically just as statistically ridiculous because there are so few Terrans in Proleague and a huge number of the Terran wins from only a couple players so it says basically nothing about TvP statistically.


Yeah, I agree completely, their methods of measuring skill just gets entangled with balance changes to the point its impossible to determine if a player is better because of his skill or race, and its hard to even measure the impact changes can have.

I think they need to look a combination of win rates, race distribution on ladder and in tournaments, race representation deep into tournaments as well as just some old fashioned watching as many games as possible and seeing how the game actually plays out.
WriterNever give up, never surrender! https://www.youtube.com/user/DestructiconSC
Squat
Profile Joined September 2013
Sweden7978 Posts
January 23 2014 12:30 GMT
#1156
On January 23 2014 21:05 Plansix wrote:
Show nested quote +
On January 23 2014 21:03 Bagi wrote:
On January 23 2014 21:00 Daswollvieh wrote:
And now, make it four ST players in Code S.

He still has to overcome the biggest obstacle of them all: a protoss.

You heard it hear folks, zerg under powered vs terran. Blizzard please nerf.

Are being facetious/trolling or just not reading?
"Digital. They have digital. What is digital?" - Donald J Trump
vthree
Profile Joined November 2011
Hong Kong8039 Posts
Last Edited: 2014-01-23 12:43:46
January 23 2014 12:35 GMT
#1157
On January 23 2014 21:24 Destructicon wrote:
Show nested quote +
On January 23 2014 21:16 Phobos_ wrote:
On January 23 2014 20:39 hfsrj wrote:
I know it's rude to auto-quote, but I'd like to have some comments on this :

On January 23 2014 04:55 hfsrj wrote:
Btw I don't know if they noticed that most way to "correct for skill" actually hide the imbalance, especially if the most powerfull race is more represented than the two other.

It's quite easy to show : let's say it's 25%A, 25%B, 50%C, for the sake of the argument. If C is favored, and you are A, you loose most of your games versus C, for equal skill (say 100% for simplicity, the argument is still valid for lower). Then you proceed to loose 50% of your games against A and B (same skill). You therefore lost 75% of your games, and your MMR (perceived skill) goes down untill you reach 50% win ratio, i.e. your perceived skill is lower than actual skill.

Things are more complicated in truth, but the statistics still apply. If they count skill by MMR, the figure they show are erroneous. This phenomena exist if C is not overrepresented, but it is obviously smaller, and much harder to correct for.


Anyone knows how the skill correction is made ? It's a bit hard not to be biased in doing it, so I'd really like to know more. I'm a physicist and quite comfortable with statistics, btw, and I think it'd be nice in this topic to have more arguments than "small sample size" and "corrected win percentage", that are actually not very informative.


The only way (that we know of anyway) that Blizzard can measure skill is the mmr system. So most likely what they are doing is similar to how they update the mmr of players, some kind of Bayesian update. When two players play a game the result gives us more information about the skill or mmr of each player. Bayes' rule tells us how to compute this. It is kind of like a weighting, if player A is much more skilled then player B and player A wins then the system does not change very much because the outcome was expected with high probability. However, if player B wins then that is not expected and a larger change to the players mmrs would occur.

You can do a similar thing with race match ups. If a masters Zerg plays a masters Terran then the system knows the skills of each player and so it has some kind of idea on what the expected outcome is. So if the Zerg player has a much higher mmr than the Terran and the zerg wins then the system would not consider that very good evidence that Zerg has a higher win rate than Terran at a similar skill level in the masters league, ie skill is getting factored out. You can actually make this mathematically rigorous pretty easily.

The problem with this is that imbalance in the game gets incorporated into a player's skill or mmr after they play some games. If you give marines like 10 range, after all the Terran's play a bunch of games their mmr or "skill" is going to go up because they will start beating better players. We as people know this is imbalance but the system doesn't differentiate because it can't know the difference. So basically when they say they are factoring out skill they are most likely factoring out imbalance at the same time.

If this is actually what they are doing I find it pretty comical that they use this along with Proleague results to justify anything. The results from Proleague are basically just as statistically ridiculous because there are so few Terrans in Proleague and a huge number of the Terran wins from only a couple players so it says basically nothing about TvP statistically.


Yeah, I agree completely, their methods of measuring skill just gets entangled with balance changes to the point its impossible to determine if a player is better because of his skill or race, and its hard to even measure the impact changes can have.

I think they need to look a combination of win rates, race distribution on ladder and in tournaments, race representation deep into tournaments as well as just some old fashioned watching as many games as possible and seeing how the game actually plays out.


I am pretty sure David Kim just watches PL atm... One week after Ts do well, he posts. He had been silent for so long and since. Basically, he cherry picks... Honestly, if he wants to comment on balance, he should have a regular schedule to talk about state of the game and balance (every 2 weeks)?

Honestly, he should have a schedule where he posts on his balance thoughts because then everyone would get a fair treatment. He shouldn't post just because the recent data supports his decisions.

Btw, David Kim, what are you seeing in Code A right now? I would like to see you make a post on that? Maybe you should teach the Code A terran what they are doing wrong? I am sure they just aren't getting it... Meanwhile, I would suggest making all Code S PvZ games played on Daedulus since that is the map I enjoy most as a Terran, I want to see how Protoss pro will adapt.
deacon.frost
Profile Joined February 2013
Czech Republic12129 Posts
January 23 2014 12:46 GMT
#1158
On January 23 2014 21:35 vthree wrote:
Show nested quote +
On January 23 2014 21:24 Destructicon wrote:
On January 23 2014 21:16 Phobos_ wrote:
On January 23 2014 20:39 hfsrj wrote:
I know it's rude to auto-quote, but I'd like to have some comments on this :

On January 23 2014 04:55 hfsrj wrote:
Btw I don't know if they noticed that most way to "correct for skill" actually hide the imbalance, especially if the most powerfull race is more represented than the two other.

It's quite easy to show : let's say it's 25%A, 25%B, 50%C, for the sake of the argument. If C is favored, and you are A, you loose most of your games versus C, for equal skill (say 100% for simplicity, the argument is still valid for lower). Then you proceed to loose 50% of your games against A and B (same skill). You therefore lost 75% of your games, and your MMR (perceived skill) goes down untill you reach 50% win ratio, i.e. your perceived skill is lower than actual skill.

Things are more complicated in truth, but the statistics still apply. If they count skill by MMR, the figure they show are erroneous. This phenomena exist if C is not overrepresented, but it is obviously smaller, and much harder to correct for.


Anyone knows how the skill correction is made ? It's a bit hard not to be biased in doing it, so I'd really like to know more. I'm a physicist and quite comfortable with statistics, btw, and I think it'd be nice in this topic to have more arguments than "small sample size" and "corrected win percentage", that are actually not very informative.


The only way (that we know of anyway) that Blizzard can measure skill is the mmr system. So most likely what they are doing is similar to how they update the mmr of players, some kind of Bayesian update. When two players play a game the result gives us more information about the skill or mmr of each player. Bayes' rule tells us how to compute this. It is kind of like a weighting, if player A is much more skilled then player B and player A wins then the system does not change very much because the outcome was expected with high probability. However, if player B wins then that is not expected and a larger change to the players mmrs would occur.

You can do a similar thing with race match ups. If a masters Zerg plays a masters Terran then the system knows the skills of each player and so it has some kind of idea on what the expected outcome is. So if the Zerg player has a much higher mmr than the Terran and the zerg wins then the system would not consider that very good evidence that Zerg has a higher win rate than Terran at a similar skill level in the masters league, ie skill is getting factored out. You can actually make this mathematically rigorous pretty easily.

The problem with this is that imbalance in the game gets incorporated into a player's skill or mmr after they play some games. If you give marines like 10 range, after all the Terran's play a bunch of games their mmr or "skill" is going to go up because they will start beating better players. We as people know this is imbalance but the system doesn't differentiate because it can't know the difference. So basically when they say they are factoring out skill they are most likely factoring out imbalance at the same time.

If this is actually what they are doing I find it pretty comical that they use this along with Proleague results to justify anything. The results from Proleague are basically just as statistically ridiculous because there are so few Terrans in Proleague and a huge number of the Terran wins from only a couple players so it says basically nothing about TvP statistically.


Yeah, I agree completely, their methods of measuring skill just gets entangled with balance changes to the point its impossible to determine if a player is better because of his skill or race, and its hard to even measure the impact changes can have.

I think they need to look a combination of win rates, race distribution on ladder and in tournaments, race representation deep into tournaments as well as just some old fashioned watching as many games as possible and seeing how the game actually plays out.


I am pretty sure David Kim just watches PL atm... One week after Ts do well, he posts. He had been silent for so long and since. Basically, he cherry picks... Honestly, if he wants to comment on balance, he should have a regular schedule to talk about state of the game and balance (every 2 weeks)?

Honestly, he should have a schedule where he posts on his balance thoughts because then everyone would get a fair treatment. He shouldn't post just because the recent data supports his decisions.

Btw, David Kim, what are you seeing in Code A right now? I would like to see you make a post on that? Maybe you should teach the Code A terran what they are doing wrong? I am sure they just aren't getting it... Meanwhile, I would suggest making all Code S PvZ games played on Daedulus since that is the map I enjoy most as a Terran, I want to see how Protoss pro will adapt.

For the start - don't build vikings to counter hallucinated colossus?
I imagine France should be able to take this unless Lilbow is busy practicing for Starcraft III. | KadaverBB is my fairy ban mother.
SeXyBaCk
Profile Joined January 2012
Switzerland174 Posts
Last Edited: 2014-01-23 12:53:09
January 23 2014 12:51 GMT
#1159
Anyone who refuses to see that TvP is completely broken now is delusional. Pro-league means nothing. I can't remember the last time a terran even remotely won a set easily against protoss. Winning usually involves 3 pronged attacking and a protoss walking their colossus or templars into a small unit of bio and losing the game due to poor control. Whenever there's army clashing terran dies.

This has simply happened because one by one every terran timing has been nerfed out of the game. What do you want to achieve playing TvP these days? What are you going for? What's the aim? Deny a third? Deny a 4th? Get to lategame ghost viking? What? 2-2 pushes? None of it remotely works. All it is now is surviving whatever massive attack protoss throws at you and then stay in the game hoping protoss will somehow mess up their control and throw away their aoe damage.

I don't think TvZ has been played enough yet post mine nerf to call it balanced. There seems to be way more mech play because the mines feel close to useless against mutas these days, as well as the hits vs bane ling feel completely random. Zerg have figured out mines, and even if the nerf was undone I think they could still handle them and results would be the same.

Now I understand balance team addressing mines, at times they felt devestating and change a game within 1 second. Why TvP hasn't been addressed in 2 years of protoss dominating the matchup is utterly beyond me.

Someone tell me if were to bet money on starcraft between pro players, would anyone right now bet on the terran when up against the protoss if you knew nothing about the players? Are you putting your money on the protoss or the terran? Be honest.

That is not balance. It's not even remotely close. 95% will honestly put their money on the protoss.
dargul
Profile Joined May 2010
Russian Federation125 Posts
January 23 2014 12:52 GMT
#1160
So now we have final stats regarding TvP in Cod A
Terrans won only 1 bo3 and lost 15...
If we are talking about sets: 32-7 only 18% winrate.

Tomorrow 2terrans will have a chance to get into code S simply because there are no toss in that group.
And by the way BByong and SuperNova got in Code S only because they have won Zerg or Terran players.

So we gonna have 3-5т terrans in Code S out of 32 it is 9-15% of terrans.

Nice job D Kim.
In Stim We Trust
Prev 1 56 57 58 59 60 62 Next
Please log in or register to reply.
Live Events Refresh
Kung Fu Cup
11:00
#5
RotterdaM341
TKL 130
IndyStarCraft 75
SteadfastSC74
Rex69
Liquipedia
Replay Cast
09:00
KungFu Cup 2026 Week 4
LiquipediaDiscussion
[ Submit Event ]
Live Streams
Refresh
StarCraft 2
RotterdaM 341
Lowko154
TKL 130
IndyStarCraft 75
SteadfastSC 74
Rex 69
ProTech23
StarCraft: Brood War
Britney 24579
Horang2 2090
Bisu 1174
Jaedong 908
Hyuk 728
firebathero 685
Soma 368
Zeus 366
Mini 276
actioN 251
[ Show more ]
Stork 203
Pusan 196
Snow 165
Larva 161
Last 149
ZerO 137
Light 135
PianO 124
hero 90
Soulkey 84
Rush 82
Hyun 70
sorry 49
Hm[arnc] 45
[sc1f]eonzerg 43
Sharp 35
NaDa 30
Shinee 26
Sacsri 24
Bale 22
soO 21
Barracks 19
Free 19
HiyA 12
Terrorterran 12
Icarus 11
Movie 8
ajuk12(nOOB) 7
Noble 4
ggaemo 1
Dota 2
XaKoH 679
canceldota177
League of Legends
Reynor46
Counter-Strike
x6flipin425
allub177
markeloff47
Other Games
singsing1611
B2W.Neo994
DeMusliM226
Pyrionflax216
Mew2King75
QueenE15
ZerO(Twitch)8
Organizations
Other Games
BasetradeTV1670
Counter-Strike
PGL185
StarCraft 2
Blizzard YouTube
StarCraft: Brood War
BSLTrovo
sctven
[ Show 13 non-featured ]
StarCraft 2
• CranKy Ducklings SOOP23
• AfreecaTV YouTube
• intothetv
• Kozan
• IndyKCrew
• LaughNgamezSOOP
• Migwel
• sooper7s
StarCraft: Brood War
• BSLYoutube
• STPLYoutube
• ZZZeroYoutube
League of Legends
• Jankos1210
• TFBlade1108
Upcoming Events
Replay Cast
12h 20m
The PondCast
22h 20m
WardiTV Map Contest Tou…
23h 20m
CranKy Ducklings
1d 12h
Escore
1d 22h
WardiTV Map Contest Tou…
1d 23h
OSC
2 days
Korean StarCraft League
2 days
CranKy Ducklings
2 days
WardiTV Map Contest Tou…
2 days
[ Show More ]
IPSL
3 days
WolFix vs nOmaD
dxtr13 vs Razz
BSL
3 days
Sparkling Tuna Cup
3 days
WardiTV Map Contest Tou…
3 days
Ladder Legends
4 days
BSL
4 days
IPSL
4 days
JDConan vs TBD
Aegong vs rasowy
Replay Cast
4 days
Replay Cast
4 days
Wardi Open
4 days
Afreeca Starleague
4 days
Bisu vs Ample
Jaedong vs Flash
Monday Night Weeklies
5 days
RSL Revival
5 days
Afreeca Starleague
5 days
Barracks vs Leta
Royal vs Light
WardiTV Map Contest Tou…
5 days
RSL Revival
6 days
Liquipedia Results

Completed

Proleague 2026-04-13
RSL Revival: Season 4
NationLESS Cup

Ongoing

BSL Season 22
ASL Season 21
CSL 2026 SPRING (S20)
IPSL Spring 2026
StarCraft2 Community Team League 2026 Spring
Nations Cup 2026
IEM Rio 2026
PGL Bucharest 2026
Stake Ranked Episode 1
BLAST Open Spring 2026
ESL Pro League S23 Finals
ESL Pro League S23 Stage 1&2
PGL Cluj-Napoca 2026
IEM Kraków 2026

Upcoming

Escore Tournament S2: W3
Acropolis #4
BSL 22 Non-Korean Championship
CSLAN 4
Kung Fu Cup 2026 Grand Finals
HSC XXIX
uThermal 2v2 2026 Main Event
RSL Revival: Season 5
2026 GSL S1
WardiTV TLMC #16
IEM Cologne Major 2026
Stake Ranked Episode 2
CS Asia Championships 2026
Asian Champions League 2026
IEM Atlanta 2026
PGL Astana 2026
BLAST Rivals Spring 2026
CCT Season 3 Global Finals
TLPD

1. ByuN
2. TY
3. Dark
4. Solar
5. Stats
6. Nerchio
7. sOs
8. soO
9. INnoVation
10. Elazer
1. Rain
2. Flash
3. EffOrt
4. Last
5. Bisu
6. Soulkey
7. Mini
8. Sharp
Sidebar Settings...

Advertising | Privacy Policy | Terms Of Use | Contact Us

Original banner artwork: Jim Warren
The contents of this webpage are copyright © 2026 TLnet. All Rights Reserved.