EPTDesignated Balance Discussion Thread - Page 789
| Forum Index > SC2 General |
|
RaFox17
Finland4581 Posts
| ||
|
Big J
Austria16289 Posts
On November 21 2013 06:39 Hider wrote: Let's say a nerf to terran occurs. Even though we in theory thinks that it is someting that only should impact higher level play, it will still cause more terrans to be placed in bronze relative to the other races. After the nerf is implemented, the following till happen on the ladder; Ex low master terrans --> Demoted to diamond league --> Ex low diamonds --> Demoted to plat --> ex silver players goes into bronze while ex bronze protoss and zergs gets promoted to silver. Yes, and which amount of the overrepresentation in bronze would this explain? We don't know! It could be that all the 2000 guys come from that, it could be that none of them can be explained like that, since we don't know how the distributions would look like in a balanced scenario. You ask: Since that is exactly what the data has looked like on a very consistent basis, it seems reasonable to assume that is the truth unless we can proove otherwise/come up with strong arguments for why that is not the case. You answer yourself: Because it has consistently been like that. So unless you want to say that Terran has been underpowered in 2010/11 as well or at the start of HotS (because in those time frames, Terran had more representation in the lower leagues than in the higher ones too), your argument does not hold for all times. The question is whether the statistic can be bias'ed enough for there to be another explanation than terran being UP. I pointed to two forms of bias; One of them was in the direction of favouring zerg, the other one favoured terran. So I made a rough estimate and assumed they cancelled out each other. You made two assumptions, neither of which can be (dis-)proven. Even more without arguing why they should be the only ones or what weight they have in the whole picture. I don't blame you, because it's not possible to argue. That's why I originally said: The distributions themselve are worthless. People keep on saying that "winrates aren't everything", the plain race distributions are much worse, since we don't know the reasons why people play which race at what level. You simply cannot get a reasonable precision in your model that explains +/- 2000 players at bronze of (45000 Terrans) and we simply cannot know whether they are outliers, coming from some impacts, or whether they have been shifted down by balance. It's far too few of them to even explain them with a top-model.* *even our very best, specialized, statistical forecast tools are on average 8.5% off, so I don't believe that you can make a prognosis how the perfect Starcraft race distributions should look like (assuming perfect balance), just based of arguments. | ||
|
InfMorse
United States16 Posts
http://www.teamliquid.net/forum/viewmessage.php?topic_id=436143 | ||
|
Chaggi
Korea (South)1936 Posts
On November 21 2013 15:36 RaFox17 wrote: Couldn´t you but the recent toss success down to the fact that Dear and SoS just might be freaking good players? Maybe it´s because they are really, really good that they won? When Innovation was wrecking everyone people said that it was because he was so good and even said that you can´t use him in balance arguments cause he skewed the stats with his godlike-skill. TvP might be difficult at the moment for terran and yes toss has million and hundred all-ins but maybe we should give Dear and SoS some respect they deserve. The problem is figuring out where is that skill, and where that's because the race is OP. BW was unique cause the mechanical skill required to do anything in that game was so high, that that's how skill was really defined for awhile. If you look at SC2 as well, you might have two really evenly matched players in "skill", but if it's a CC first on the low ground vs a 6 pool, it's a straight up build order loss and the 6 pooler is gonna win everytime, and not just are they gonna win everytime, the CC first guy is gonna look absolutely stupid doing whatever build he was doing. Not to take anything away from Dear or sOs either, they're absolutely amazing players, and should be considered just as good if not better than anyone else in the world (on that day). | ||
|
ejozl
Denmark3392 Posts
The problem is figuring out where is that skill, and where that's because the race is OP. BW was unique cause the mechanical skill required to do anything in that game was so high, that that's how skill was really defined for awhile. If you look at SC2 as well, you might have two really evenly matched players in "skill", but if it's a CC first on the low ground vs a 6 pool, it's a straight up build order loss and the 6 pooler is gonna win everytime, and not just are they gonna win everytime, the CC first guy is gonna look absolutely stupid doing whatever build he was doing. Not to take anything away from Dear or sOs either, they're absolutely amazing players, and should be considered just as good if not better than anyone else in the world (on that day). There was a lot of early game cheese in BW too. | ||
|
Chaggi
Korea (South)1936 Posts
On November 21 2013 17:51 ejozl wrote: There was a lot of early game cheese in BW too. That was a bad example. I wanted to really say that if two players of equal skill are playing, and one has a worse build, they can lose in ways that make them look like a diamond leaguer | ||
|
Sabu113
United States11048 Posts
On November 21 2013 17:21 Chaggi wrote: The problem is figuring out where is that skill, and where that's because the race is OP. BW was unique cause the mechanical skill required to do anything in that game was so high, that that's how skill was really defined for awhile. If you look at SC2 as well, you might have two really evenly matched players in "skill", but if it's a CC first on the low ground vs a 6 pool, it's a straight up build order loss and the 6 pooler is gonna win everytime, and not just are they gonna win everytime, the CC first guy is gonna look absolutely stupid doing whatever build he was doing. Not to take anything away from Dear or sOs either, they're absolutely amazing players, and should be considered just as good if not better than anyone else in the world (on that day). I think he's still got a good point that it's a bit silly that all these arguments are suddenly showing up when once in a blue moon protoss players are posting tournament wins and Dear is frankly playing out of his mind. I never watched enough Innovation to buy into the hype but seeing Dear's play- he's next gen protoss with his decisions. Silly timing. Good fundamental point though.These mechanical edges are either too easy so that they are balanced against (infinite kiting) or non-existent so a lot of the games come down to errors on one side rather than ridiculous brilliance. Ofc then we get back to the same old design quibbles. | ||
|
Hider
Denmark9390 Posts
Yes, and which amount of the overrepresentation in bronze would this explain? We don't know! It could be that all the 2000 guys come from that, it could be that none of them can be explained like that, since we don't know how the distributions would look like in a balanced scenario. Well in most cases the distribution in a balanced scenario should have a somewhat equal distribution. There can ofc be scenarios (as discussed) where it isn't the case. But in the base-case we need to assume the distribution should be equal. If we can point to specific forms of bias's which impacts the optimal distribution, we then needs to make subjective adjustments. My question to you is what kind of bias's do you see that could impact the balanced distribution (?) You answer yourself: Because it has consistently been like that. So unless you want to say that Terran has been underpowered in 2010/11 as well or at the start of HotS (because in those time frames, Terran had more representation in the lower leagues than in the higher ones too), your argument does not hold for all times. Terran wasn't always underrepresented in diamond league/master league/GM. In fact, it especially in Korea had periods where I remember the reverse was the case (like 40% of all GM's were terrans). However, it has been like that since mid 2011 which fits well into my theory of the ghost nerf + infestor buff nerfing terran too much. You simply cannot get a reasonable precision in your model that explains +/- 2000 players at bronze of (45000 Terrans) and we simply cannot know whether they are outliers, coming from some impacts, or whether they have been shifted down by balance. It's far too few of them to even explain them with a top-model.* Here is one model. These numbers are created so they will create a somewhat equal distribution 300 high-dia terrans should be in masters league if the game was balanced. 600 high-plat terrans should be in dia if the game was balanced 900 high gold terrans should be in plat if the game was balanced 2000 high silver terrans should be in gold if the game was balanced 3000 high bronze terrans should be in silver if the game was balanced The numbers are btw higher in the lower leagues are the amount of players is higher. But I don't see how this model can't possibly be the case. even our very best, specialized, statistical forecast tools are on average 8.5% off, so I don't believe that you can make a prognosis how the perfect Starcraft race distributions should look like (assuming perfect balance), just based of arguments What does forecasting tools have to do with anything here? Here we are looking at historical data (we aren't forecasting anything). We can see that the data fits well into the hypothesis that terran became UP at some point in 2011. That makes sense as the nerfs it faced at that point were severe. Due to the large sample size and the large disceprencancy between equal and actual distribution, the differences are statistical significant. So I wonder, is there any other other reasonable expalantion for why terrans distribution is skewed towards bronze that fits the data well? | ||
|
Hider
Denmark9390 Posts
On November 21 2013 11:38 Wingblade wrote: And now your guilty of a straw manat the top of your comment. You have a great ability to use stats to your opinion, but your construction of logical arguments is awful. At no point did I ever explicitly state or even imply that tournaments are the only factor in arguing balance. I was responding to you drawing conclusions(again of a false dichotomy nature) from only ladder representation. So each of your three points are inherently flawed, because they are counters to an argument that I never presented, but that you made up in place of my argument. In terms of proving that the tournament win rates and information are "statistically significant", look back to the days when Terran was considered OP. Both ladder representation AND tournament results were skewed in favor of Terran. Likewise, at the end of WoL, both matched the belief in the communist that Zerg was imbalanced. Right now that isn't the case. Yes, Dear won two tournaments, playing better than his opponents. Yes, sOs won Blizzcon, by being ahead of the metagame and playing fantastically well in his own right. Three tournaments do not create the correlation needed, especially when everything else does not support that(in regards to tournaments). When there has been a clear imbalance in SC2 history, both ladder representation and tournament results supported the same conclusion. That is not the case right now. Any attempt to use one or the other exclusively is erroneous. Anyways, we just had a new major patch go through, which means that we can't possibly know about balance this early into the metagame development. The problem is that it simple is too easy to tell a few stories based on low sample sizes. If we are looking isolated at GSL, the sample size is way too low and using it for anything other than qualitaitve analysis of highest game caliber is a dumb idea. If we combine all (international) tournaments data - are you even sure that the data fits your hypothesis? Foreign terran players haven't really won anything for since 2010. But even tournament wins on a large basis can be hugely impacted by whether a foreign team (TL or EG) picked up a terran or protoss player during 2011-2012. For instance, if EG had not picked up Puma but a protoss player, terrans tournamnet results would have been alot worse in 2011. Now if TL had not picked up Hero but a zerg player, protoss results would be even worse. Point is, that these types of decisions by foreign management has little to do with balance, but impacts tourmanet win percetnage by a large margin. That is why looking only at tournament wins - regardless of sample size - is also a dumb idea. In fact, I could present the reverse argument: Demuslim is a ladder hero, but he gets abused in tournaments because opponents can do 100 different things against him while terrans options are limited. In my opinion these types of stories are simply useless in quantitative analysis. So what I asked your for was not a couple of stories or examples - but a large amount of tournament game-by-game data which showed that protoss performs better on the ladder than in tournaments. Doing that is ofc a very time-consuming, but you simply cannot go around saying that the protoss is UP in tournaments without having provided robust arguments. At this point you are simply providing a theory. | ||
|
Deleted User 137586
7859 Posts
On November 21 2013 11:12 Wingblade wrote: Yea, Dear and sOs. Name any other Protoss who have had success in the last 3 to 4 months and sustained it for any period of time. Hell sOs had 1 weekend of success that could very well be a flash in the pan. "Completely wrecked" is an awfully bold statement. Care to back up your hyperbole with any actual evidence? First, it's not hyperbole, it's facts that the last 3 most important tournaments have been won by Protoss. For other consistent Protoss performers: Naniwa (he might have lost against Soulkey, but, hey, it's bloody Soulkey). DuckDeok refused to back down after winning his first tournament, recently even crushing Innovation. Everyone forgets that players such as Zest, First, Parting, Rain all have consistent excellent results at the hardest level of competition (although they might not win as many tournaments as they focus on team leagues and GSL - another reason why looking at tournament winners is a bad idea). As for recent qualifiers, recall Red Bull and IEM Season VIII. Protoss had twice the number qualify as the other races. Also recall Sora's recent rise to fame. Also Aligulac says that there's a slight trend towards P>T (although it's minor). So it's ridiculous to say that Protoss is bloody well underpowered at tournaments. I cannot believe that anyone is so deluded as to make that claim. As for the point that "the timing for balance discussion is weird as Protoss only recently won a tournament". The timing has more to do with several nerfs to terran in the MU (hellbat nerf made them a ForGG only unit, WM nerf reduced the viability of WM drops and oracle harass is now more potent). Of course terrans will raise the issue of balance, if the balance is being changed by balance patches. | ||
|
Hider
Denmark9390 Posts
On November 21 2013 16:43 InfMorse wrote: My thread was shut down cause it was considered a balance discussion. All though my point wasn't to argue balance but to better understand it's measurements it was locked. However, I am still unsatisfied with my understanding. Could one or 5 of you look this over and help me understand how Blizzard can say the game is balanced when the international race distribution according to leagues is so off? http://www.teamliquid.net/forum/viewmessage.php?topic_id=436143 I wrote this a couple of pages ago; Btw, while Blizzard does calculate and adjusted skill-based W/R, chances are that it doesn't look at the distribution in a proper way. Look at the below video from October 2010; http://www.teamliquid.net/forum/viewmessage.php?topic_id=250789 Notice that when we go from unadjusted to adjusted, these changes occur; - TvZ w/r is unchanged - PvZ goes up by 2% in favor of protoss - PvT w/r goes up by 7% in favor of protoss So in the adjustment data, the skill-indicator goes up alot for terran in TvP. In PvZ it goes marginally up for zerg. So basically Blizzard believes protoss raw ladder data didn't fully take into account how strong the race actually was. But how on earth can they get to that conclusion. The ladder distribution was similar to that of protoss and terran right? Further, the fact that there were a lot fewer zerg players and that it is likely that there was some bias amongst lesser skilled players to pick terran over zerg, wasn't taken into account, since the skill-adjustmented in TvZ was unchanged. The only skill-adjustmented proces I can see that makes a minimal amount of sense is for blizzard to assume that those who play the game more should be weaker than those who practice less. So if a terran player has played 10K minutes and a protoss player has played 8K minutes, and they were of equal skill when they started playing the game and still are equally skilled today --> Protoss is OP. That's, however, a much less efficient adjustment proces than just relying on an estimated w/r based on "the balanced ladder distribution". I think that Blizzard probably has a couple of statisticans/mathematicians hired. When Sc2 was developed he was helping the Sc2 coming up with a skill-adjusted formula that was relevant when the game was released. Possibly it could look something like this: Core MMR is decided after the first 30 games. There is no skill-adjusted proces untill you have gotton your core MMR. Let's say you are a terran player and you get placed into platinum after having reached your core MMR. On 1 season you have played 500 gamesand your MMR has increased by 10%. However, that is below that of an equally skilled protoss player that played 500 games as they have improved by 20%. Since you have played a higher number of games your skill-adjusted MMR is higher than your actual MMR, while the reverse is the case for the protoss player. Its possibly that there also are other factors which impacts the skill-based MMR, such as a given a lower weight to teamgames and an even lower weight to custom games. I find it unlikely that metrics such as SQ or EAPM as they very flawed in terms of providing a measure of skill. Blizzard's approach was probably a decent idea when the game was released, but I believe that it today is almost useless and that Blizzard should use an estimated probably of winning based on distribution rather than "skill-based". But the mathematician is probably busy working on WOW, Titan or Hearthstone etc. | ||
|
RParks42
United States77 Posts
| ||
|
Green_25
Great Britain696 Posts
On November 21 2013 15:36 RaFox17 wrote: Couldn´t you but the recent toss success down to the fact that Dear and SoS just might be freaking good players? Maybe it´s because they are really, really good that they won? When Innovation was wrecking everyone people said that it was because he was so good and even said that you can´t use him in balance arguments cause he skewed the stats with his godlike-skill. TvP might be difficult at the moment for terran and yes toss has million and hundred all-ins but maybe we should give Dear and SoS some respect they deserve. Yeah, I think this is mostly the case. Dear and SoS are both just freaking good, though SoS is more mind gamey. Basically, macro play is say... 5% stronger now for Protoss due to the Mothership Core, otherwise Protoss is mostly the same. Stargate/blink all ins are SUPER annoying, but they aren't broken, just really tough to deal with. At the highest level you don't see stargate all ins as much, the best Protoss are using oracles for macro play now. | ||
|
Hider
Denmark9390 Posts
I used a sample size of 690 of the lowest ranked global bronze players and found that 32.7% of them were terran players. Now in average 33.1% of all terran players are bronze league, which didn't give any indicaiton at all for there to be some kind of truth that theory. This just further backs up the "terran up"-theory as the most likely explanation. | ||
|
Chaggi
Korea (South)1936 Posts
On November 21 2013 18:21 Sabu113 wrote: I think he's still got a good point that it's a bit silly that all these arguments are suddenly showing up when once in a blue moon protoss players are posting tournament wins and Dear is frankly playing out of his mind. I never watched enough Innovation to buy into the hype but seeing Dear's play- he's next gen protoss with his decisions. Silly timing. Good fundamental point though.These mechanical edges are either too easy so that they are balanced against (infinite kiting) or non-existent so a lot of the games come down to errors on one side rather than ridiculous brilliance. Ofc then we get back to the same old design quibbles. It's a bit sad IMO. Dear has some super crisp timings and super amazing play that even I, a Terran player, can recognize. Innovation had some really good showings pre HoTS even though that's where he really shined. I didn't pay that much attention to him at first, but when one of my friends told me he's the perfect Terran, I started to watch his games more. I think my friend was like, he's got better micro than MKP, better macro than Bomber and had the best timings in the world. When GSL was dominated by Zerg, it was really refreshing to see someone new pop up. | ||
|
-Celestial-
United Kingdom3867 Posts
On November 21 2013 12:03 Courthead wrote: In the last 3 or 4 months Protoss players have been absolutely dominating, Protoss domination in the last three or four months? Ok then. Lets look at the highest level tournaments from the last four months (source: Liquidpedia's list of Premier Tournaments): ASUS ROG Summer 2013 - Won by Taeja (Terran), Second place San (Protoss) WCS Season 3 Korea - Won by Dear (Protoss), Second place soO (Zerg) WCS Season 2 Finals - Won by Bomber (Terran), Second place Jaedong (Zerg) WCS Season 3 Europe - Won by MMA (Terran), Second place MC (Protoss) DreamHack Open: Bucharest - Won by Taeja (Terran), Second place INnoVation (Terran) WCS Season 3 America - Won by Polt (Terran), Second place ByuL (Zerg) IEM Season VIII - New York - won by Life (Zerg), Second place NaNiwa (Protoss) WCS Season 3 Finals - Won by Dear (Protoss), Second place Soulkey (Zerg) WCS Global Finals - Won by sOs (Protoss), Second place Jaedong (Zerg) HomeStory Cup VIII - Won by Taeja (Terran), Second place HyuN (Zerg) Totals: Terran - 6 first place, 1 runner up (7 total finalists) Protoss - 3 first place, 3 runner up (6 total finalists) Zerg - 1 first place, 6 runner up (7 total finalists) Yeah check out that utter domination by Protoss! Oh wait...half the actual finals wins compared to Terran and roughly even finalist numbers. Looks more or less even, maybe a slight Terran advantage in longer series (finals are usually longer series and Terran are winning more finals than expected) but I'd hesitate to draw anything from just 5 non-mirror tournament finals. Seriously, its language like "dominating" when it isn't backed up by facts that makes these arguments utterly comical and get nobody anywhere... Although fun fact I realised from checking that data out: Terran hasn't lost a Premier tournament finals they've been in against another race since March. And have only lost two in the whole of HotS. | ||
|
Deleted User 137586
7859 Posts
Do not count who gets random wins across several weird tournaments, look at actual tournament performance at the highest level. In GSL, the top 8 had 4 protoss, definitely a healthy number. The tournament was won by Dear, although my money was on Trap taking it. I'd say the game was pretty balanced in the last few months, although P was definitely coming into a new era. The weird thing is that Blizz intervened and made P stronger. - and people are saying that the current balance does not warrant this change, and risks becoming another queen patch. | ||
|
NarutO
Germany18839 Posts
On November 21 2013 11:12 Wingblade wrote: Yea, Dear and sOs. Name any other Protoss who have had success in the last 3 to 4 months and sustained it for any period of time. Hell sOs had 1 weekend of success that could very well be a flash in the pan. "Completely wrecked" is an awfully bold statement. Care to back up your hyperbole with any actual evidence? Hey man, whats your opinion on the post below? On November 21 2013 09:36 TheDwf wrote: Complete nonsense. Quite on the contrary, the fact Protoss has so many viable builds/styles is a strength in tournaments because it allows them to vary their play. Rain's defensive play is a stylistic choice, he does not have to play like that 100% of the time since his race has countless agressive options; thus his predictability cannot be blamed on his race alone. PartinG relied on his Immortal all-in because Zerg was broken in WoL, zero news here; if he can no longer wins PvZ now, the issue is on his end since other top Protoss players don't have this problem. If MC's relative fall is to be blamed on his tendancy to play agressive/all-in too much (which is a simplistic explanation, but whatever), it's again his fault because his race has passive macro options; thus his predictability cannot be blamed on his race alone, bis. The very examples you provide contradict your own thesis. | ||
|
Big J
Austria16289 Posts
Basically everything you say is built upon certain assumptions, that I simply don't support: Well in most cases the distribution in a balanced scenario should have a somewhat equal distribution. No, that assumes that the player numbers would be 33-33-33 (-random) in perfect balance. If we can point to specific forms of bias's which impacts the optimal distribution, we then needs to make subjective adjustments. My question to you is what kind of bias's do you see that could impact the balanced distribution (?) Yes, that's exactly the problem. I can name a thousand reasons what could or could not influence the amount of players for each race. I don't, because we have no way to weight them. It's simply not easily possible and therefore just blindguessing. Terran wasn't always underrepresented in diamond league/master league/GM. In fact, it especially in Korea had periods where I remember the reverse was the case (like 40% of all GM's were terrans). Yes, but as far as I remember it has always been overrepresented in bronze. And in 2010/11 I'm pretty sure that this wasn't balance-introduced, which then questions why it has to be balance-introduced now. Not to mention that many, many statistics - pro-winrates, tournament results - show no sign of imbalanced against Terran in HotS 2013, so the numbers should have changed to a better Terran representation in the higher leagues in your theory, compared to end of WoL. Have they? The numbers are btw higher in the lower leagues are the amount of players is higher. But I don't see how this model can't possibly be the case. I'm not saying it can't be the case. I'm saying we have no way to prove it. I can also give you a model: Terrans gets chosen by 27%* of all players as their "final race choice" and the game has been balanced for a longer periode of time now, so we have reached a stable state in which everyone is in the league where he belongs. Then the number should look like this: 1270 (Master) 2228 (Diamond) 3413 (Platinum) 8345 (Gold) 12019 (Silver) 14969 (Bronze) Real numbers as comparison 1,244 2,264 3,491 8,162 12,550 17,432 *based upon the median percentage of Terran players per league. Addendum to the theory: There has been an Event X at the beginning that led to an inflation of beginners for Terran (e.g. WoL campaign; e.g. Terran being most similar to other RTS game races; e.g. preview/advertisment videos; e.g. Battlereports; e.g. Forums...) and a certain amount of players has never reconsidered their choice (and of those who have 27% still stayed Terran). I don't see why your theory would be any better than mine. Not that I think mine has to be true. As I said before, it's guessing. It's as unproveable as yours. What does forecasting tools have to do with anything here? Here we are looking at historical data (we aren't forecasting anything). We can see that the data fits well into the hypothesis that terran became UP at some point in 2011. That makes sense as the nerfs it faced at that point were severe. Due to the large sample size and the large disceprencancy between equal and actual distribution, the differences are statistical significant. Your whole argument is based upon "the distributions being even in perfect balance", which - since according to you balance isn't perfect - is an estimator for the value "perfect balance". | ||
|
-Celestial-
United Kingdom3867 Posts
On November 21 2013 21:22 Ghanburighan wrote: While I don't support "protoss dominating" language, You do realise that was literally the start, middle and end of my entire point? -_- People get hypebolic as hell and throw an absolute fit when Protoss actually wins something and call it domination. Where were all these people when it was GOMTvT back in 2011? Way too many emotions going around imo. | ||
| ||
