|
An actual trend that is clear in the data is that the sample sizes are increasing. This means that Foreigners are playing against Koreans a lot more often in 2012 than last year. In 2010 there were only ~130 matches between Foreigners and Koreans (in tournaments that meet the criteria of the study). In 2012 we have reached more than 800 games between Foreigners and Koreans in Q2 2012, largely as a result of the growth of tournaments in the Foreign scene and the Korean players travelling abroad to participate in these tournaments.
|
Still infinitely much better than in the BW days. 
|
On September 09 2012 01:54 revel8 wrote:Show nested quote +On September 09 2012 01:40 Demonhunter04 wrote:On September 07 2012 10:27 TAMinator wrote:
The only thing I find interesting is why foreigner had a big increase in win rate @ 2012 Q2. Everything else is to be expected At a glance that can probably be attributed to the queen range buff and circumstances where foreign zergs beat korean terrans/protosses more often than the reverse. No, it is just that Q2 2012 had an increase in foreigner win rate. We are still in Q3 2012 so it is too soon to know whether there is a genuine trend or it is just a blip. The OP has jumped the gun in presenting the data for Q3 when it is incomplete and thus a potential distortion.
Ah ok. I was wondering why Q3 was already there.
|
On September 09 2012 01:59 Holgerius wrote:Still infinitely much better than in the BW days.
But that's because foreigner interest was quite low. It's kinda similiar, when BW was 1/2base mostly and SC2 was 1/2 base mostly foreigners stood a chance, now it's quite bad.
And how exactly do you compare winrates in BW? You can only kinda count 2000-2003 and during that period foreigners did good..
|
On September 09 2012 01:59 Holgerius wrote:Still infinitely much better than in the BW days.
SC2 is young, give it more time :p
|
Now take Stephano, Nerchio, and Naniwa out of there. Might be close to zero percent all around haha
|
Q2 2013 10% map wins vs kespa players.
|
On September 08 2012 23:26 .syd. wrote:Funny how many people don't understand the message of the stats, which is that the winrate for foreigners are decreasing more or less constantly, not that Korean Pros are generally better at the game. Some people seem to be quite sensitive about that fact  I think one thing that we can take away from the stats is that it's mostly the better practice environment that makes Koreans better that the (lazy) foreigners. Practice is something that pays off in the long run (talent is not for example) which is exactly what we see here.
Map winrates are more or less constant since Q2 2011. If the match winrates go down, this only means longer series are played in tournaments.
Edit: It would also be interesstin to see EU vs Korea, NA vs Korea, NA vs EU.
|
On September 09 2012 01:56 theJob wrote:
There seem to be alot of people here proclaiming that the trend of foreigners’ winratio is negatively related to time. After a quick glance I’m inclined to agree, however if we are to indulge ourselves with using a technical term like “trend” then I feel like a more thorough analysis is due.
If one were to fit an OLS regression on the material it is obvious that the trend is negative. A linear regression tells us that the relationship between the win percentage of foreigners versus Korean players is declining with roughly 1,5 % per quarter. The regression holds true on a 95% significance level and has an explanationvalue of 0,55 R^2.
However after investigating the histogram further it should become obvious that if we want to fit the most accurately representative trend line to the data we have to step away from a linear relation between the two variables. When running a quadratic OLS regression we get much better results than the linear one. There is a statistically significant relation on a 99% significance level and the R^2 value is as high as 0,90. By looking at the scatterplot where the quadratic trendline is fitted I’d have to echo esfiworlds statement that foreigners winpercentage was declining rapidly in the beginning but that it has settled or staring to rise even.
So if we are to use a technical term like “trend”, then the data from esfiworld shows that foreigners performance as of late is improving.
I am not wrong s data cant be used for anything statistical at all. The op decided to exclude observations. Even if he thinks that there are less top players playing below top 16 its still an opinion and should not effect the sample size. We all know for instance that stephano have played polt in the pool play. IPL hade a pol play that was really Korean heavy . Allot of observation have been excluded so doing a osl of any sort is out of the question (not saying your analyse is wrong but the data u are using is not good).
There allot more issues with the data itself that i don't feel like rambling up. To keep it short each tournaments win ratio is repented by diffrent variabels effecting the outcome. They are not the same so i belive u cant add them up.
|
On September 09 2012 03:15 Anomi wrote:Show nested quote +On September 09 2012 01:56 theJob wrote:
There seem to be alot of people here proclaiming that the trend of foreigners’ winratio is negatively related to time. After a quick glance I’m inclined to agree, however if we are to indulge ourselves with using a technical term like “trend” then I feel like a more thorough analysis is due.
If one were to fit an OLS regression on the material it is obvious that the trend is negative. A linear regression tells us that the relationship between the win percentage of foreigners versus Korean players is declining with roughly 1,5 % per quarter. The regression holds true on a 95% significance level and has an explanationvalue of 0,55 R^2.
However after investigating the histogram further it should become obvious that if we want to fit the most accurately representative trend line to the data we have to step away from a linear relation between the two variables. When running a quadratic OLS regression we get much better results than the linear one. There is a statistically significant relation on a 99% significance level and the R^2 value is as high as 0,90. By looking at the scatterplot where the quadratic trendline is fitted I’d have to echo esfiworlds statement that foreigners winpercentage was declining rapidly in the beginning but that it has settled or staring to rise even.
So if we are to use a technical term like “trend”, then the data from esfiworld shows that foreigners performance as of late is improving. I am not wrong s data cant be used for anything statistical at all. The op decided to exclude observations. Even if he thinks that there are less top players playing below top 16 its still an opinion and should not effect the sample size. We all know for instance that stephano have played polt in the pool play. IPL hade a pol play that was really Korean heavy . Allot of observation have been excluded so doing a osl of any sort is out of the question (not saying your analyse is wrong but the data u are using is not good). There allot more issues with the data itself that i don't feel like rambling up. To keep it short each tournaments win ratio is repented by diffrent variabels effecting the outcome. They are not the same so i belive u cant add them up.
Exluding observations is fine... He clearly stated what data he used. It would be extraordinarily unreasonable to expect anyone to do this with every starcraft game ever. Any data collected is going to be a subset of the overall data. That is fine. As long as you are aware what the data you are working with actually is (which OP clearly stated), then you can do the same analysis you would do on any other set of data...
Criteria:
- Calculated based on Team Liquid listed “Premiere Events” and “Major Events” with a prize pool of at least $10,000, at least 16 competitors
- Forfeited rounds are not calculated
- Early competition rounds with more than 64 players were not counted as skill level between players is often too disparate to make useful comparisons.
And of course tournament win ratio is affected by different variables... So what? No one is trying to calculate the effect that jet lag has on korean tournament performance. Whatever the values and/or variables may be is irrelveant to the single variable OP is looking at. He didn't claim to know what caused these results, only what they are. So his analysis is totally fine.
|
On September 09 2012 03:15 Anomi wrote:Show nested quote +On September 09 2012 01:56 theJob wrote:
There seem to be alot of people here proclaiming that the trend of foreigners’ winratio is negatively related to time. After a quick glance I’m inclined to agree, however if we are to indulge ourselves with using a technical term like “trend” then I feel like a more thorough analysis is due.
If one were to fit an OLS regression on the material it is obvious that the trend is negative. A linear regression tells us that the relationship between the win percentage of foreigners versus Korean players is declining with roughly 1,5 % per quarter. The regression holds true on a 95% significance level and has an explanationvalue of 0,55 R^2.
However after investigating the histogram further it should become obvious that if we want to fit the most accurately representative trend line to the data we have to step away from a linear relation between the two variables. When running a quadratic OLS regression we get much better results than the linear one. There is a statistically significant relation on a 99% significance level and the R^2 value is as high as 0,90. By looking at the scatterplot where the quadratic trendline is fitted I’d have to echo esfiworlds statement that foreigners winpercentage was declining rapidly in the beginning but that it has settled or staring to rise even.
So if we are to use a technical term like “trend”, then the data from esfiworld shows that foreigners performance as of late is improving. I am not wrong s data cant be used for anything statistical at all. The op decided to exclude observations. Even if he thinks that there are less top players playing below top 16 its still an opinion and should not effect the sample size. We all know for instance that stephano have played polt in the pool play. IPL hade a pol play that was really Korean heavy . Allot of observation have been excluded so doing a osl of any sort is out of the question (not saying your analyse is wrong but the data u are using is not good). There allot more issues with the data itself that i don't feel like rambling up. To keep it short each tournaments win ratio is repented by diffrent variabels effecting the outcome. They are not the same so i belive u cant add them up.
My post touched on the notion that the material shows a negative trend (which it doesn't necessarily do). I never intended to prove or disprove that the sample data is appropriate for the OPs hypothesis. So yes you are right that if one were to not accept the OPs sample for whatever reason, then it’s not interesting what kind of trend there is. But again, I never touched upon that so I'm confused why you bring that up.
Your second point seems to be a bit off. As far as I understand it the OP doesn’t intend to present a comprehensive model which explains why players loose or win tournaments but rather to show whether or nor the skill gap between foreigners and Korean players are changing. For this you would simply need a time variable to represent the flow of time and a second variable representing the skill gap. Foreigner vs Korean win percentages seem to be a reasonable demonstration of differences in skill, at least to me.
|
Keep in mind that most of the Top tier players within Korea are kept in GSL, and there are still many Code S class players who never competed in foreign scene.
If you factor in the fact that competition within Korea is much tougher than other Global competitions, numbers would probably look even more grim for the foreigners.
|
On September 09 2012 03:40 theJob wrote:Show nested quote +On September 09 2012 03:15 Anomi wrote:On September 09 2012 01:56 theJob wrote:
There seem to be alot of people here proclaiming that the trend of foreigners’ winratio is negatively related to time. After a quick glance I’m inclined to agree, however if we are to indulge ourselves with using a technical term like “trend” then I feel like a more thorough analysis is due.
If one were to fit an OLS regression on the material it is obvious that the trend is negative. A linear regression tells us that the relationship between the win percentage of foreigners versus Korean players is declining with roughly 1,5 % per quarter. The regression holds true on a 95% significance level and has an explanationvalue of 0,55 R^2.
However after investigating the histogram further it should become obvious that if we want to fit the most accurately representative trend line to the data we have to step away from a linear relation between the two variables. When running a quadratic OLS regression we get much better results than the linear one. There is a statistically significant relation on a 99% significance level and the R^2 value is as high as 0,90. By looking at the scatterplot where the quadratic trendline is fitted I’d have to echo esfiworlds statement that foreigners winpercentage was declining rapidly in the beginning but that it has settled or staring to rise even.
So if we are to use a technical term like “trend”, then the data from esfiworld shows that foreigners performance as of late is improving. I am not wrong s data cant be used for anything statistical at all. The op decided to exclude observations. Even if he thinks that there are less top players playing below top 16 its still an opinion and should not effect the sample size. We all know for instance that stephano have played polt in the pool play. IPL hade a pol play that was really Korean heavy . Allot of observation have been excluded so doing a osl of any sort is out of the question (not saying your analyse is wrong but the data u are using is not good). There allot more issues with the data itself that i don't feel like rambling up. To keep it short each tournaments win ratio is repented by diffrent variabels effecting the outcome. They are not the same so i belive u cant add them up. My post touched on the notion that the material shows a negative trend (which it doesn't necessarily do). I never intended to prove or disprove that the sample data is appropriate for the OPs hypothesis. So yes you are right that if one were to not accept the OPs sample for whatever reason, then it’s not interesting what kind of trend there is. But again, I never touched upon that so I'm confused why you bring that up. Your second point seems to be a bit off. As far as I understand it the OP doesn’t intend to present a comprehensive model which explains why players loose or win tournaments but rather to show whether or nor the skill gap between foreigners and Korean players are changing. For this you would simply need a time variable to represent the flow of time and a second variable representing the skill gap. Foreigner vs Korean win percentages seem to be a reasonable demonstration of differences in skill, at least to me.
I think u have misinterpreted me. As u can see i only quoted your post since i am well aware that he sated in his post. If it was about the op i would have quoted him . What my post was directed to was your statement alone wich is allot different from the op.
" A linear regression tells us that the relationship between the win percentage of foreigners versus Korean players is declining with roughly 1,5 % per quarter. The regression holds true on a 95% significance level and has an explanationvalue of 0,55 R^2. "
Edited: I hade to redo my post again. I tought the answer above u was yours myjob. The points you brought up are all correct. Sry for the confusion. I should have been clear on that i was only discussing the statment alone and not anything else. I am self dont think win ratio by itself is a good measurement for skills
|
I this would be even more interesting if you ran a data set counting foreigners who live in korea as part of the "koreans" data group. I'd bet that a significant amount of Foreigner vs Korean wins have come from Naniwa, Huk, Sase, Thorzain and even people like Idra or Ret that have competed in korea for a fair amount of time.
|
"You only need to train for 2 hours a day to retain your mechanics at the top level"
-Grack
|
On September 07 2012 09:59 Fionn wrote:
So you're telling me that Koreans, those folk from Asia, are better than everyone else?
Shit. Just. Got. Serious.
Damn, never would've seen it coming.
It makes sense that early on when the game first came out, foreigners had a 40% win rate vs Koreans since the game wasn't really figured out. Now that the game's evolving within Korean gaming houses, it's only a matter of time that even taking a game off a no name Korean guy on the ladder will be an achievement.
inb4 top 10 KeSPA players dominate. Right now they're adjusting to the new game, but their mechanics are far more superior than eSF players.
|
On September 09 2012 03:51 Anomi wrote:Show nested quote +On September 09 2012 03:40 theJob wrote:On September 09 2012 03:15 Anomi wrote:On September 09 2012 01:56 theJob wrote:
There seem to be alot of people here proclaiming that the trend of foreigners’ winratio is negatively related to time. After a quick glance I’m inclined to agree, however if we are to indulge ourselves with using a technical term like “trend” then I feel like a more thorough analysis is due.
If one were to fit an OLS regression on the material it is obvious that the trend is negative. A linear regression tells us that the relationship between the win percentage of foreigners versus Korean players is declining with roughly 1,5 % per quarter. The regression holds true on a 95% significance level and has an explanationvalue of 0,55 R^2.
However after investigating the histogram further it should become obvious that if we want to fit the most accurately representative trend line to the data we have to step away from a linear relation between the two variables. When running a quadratic OLS regression we get much better results than the linear one. There is a statistically significant relation on a 99% significance level and the R^2 value is as high as 0,90. By looking at the scatterplot where the quadratic trendline is fitted I’d have to echo esfiworlds statement that foreigners winpercentage was declining rapidly in the beginning but that it has settled or staring to rise even.
So if we are to use a technical term like “trend”, then the data from esfiworld shows that foreigners performance as of late is improving. I am not wrong s data cant be used for anything statistical at all. The op decided to exclude observations. Even if he thinks that there are less top players playing below top 16 its still an opinion and should not effect the sample size. We all know for instance that stephano have played polt in the pool play. IPL hade a pol play that was really Korean heavy . Allot of observation have been excluded so doing a osl of any sort is out of the question (not saying your analyse is wrong but the data u are using is not good). There allot more issues with the data itself that i don't feel like rambling up. To keep it short each tournaments win ratio is repented by diffrent variabels effecting the outcome. They are not the same so i belive u cant add them up. My post touched on the notion that the material shows a negative trend (which it doesn't necessarily do). I never intended to prove or disprove that the sample data is appropriate for the OPs hypothesis. So yes you are right that if one were to not accept the OPs sample for whatever reason, then it’s not interesting what kind of trend there is. But again, I never touched upon that so I'm confused why you bring that up. Your second point seems to be a bit off. As far as I understand it the OP doesn’t intend to present a comprehensive model which explains why players loose or win tournaments but rather to show whether or nor the skill gap between foreigners and Korean players are changing. For this you would simply need a time variable to represent the flow of time and a second variable representing the skill gap. Foreigner vs Korean win percentages seem to be a reasonable demonstration of differences in skill, at least to me. I think u have misinterpreted me. As u can see i only quoted your post since i am well aware that he sated in his post. If it was about the op i would have quoted him . What my post was directed to was your statement alone wich is allot different from the op. " A linear regression tells us that the relationship between the win percentage of foreigners versus Korean players is declining with roughly 1,5 % per quarter. The regression holds true on a 95% significance level and has an explanationvalue of 0,55 R^2. "
This interpetation of the result is correct.
If we are to use a term like "trend" then the relationship (or the coefficient) between the time variable and the win percentage variable is -1,577, under a linear OLS regression. This model has a R^2 of 0,55 and holds on a 95% significanse level.
But I was claiming that a linear trend isn't optimal and that the best fitted line shows a decrease in skillgap recently.
|
Yeah but mechanics don't make that big of a difference in SC2.
|
On September 09 2012 04:28 theJob wrote:Show nested quote +On September 09 2012 03:51 Anomi wrote:On September 09 2012 03:40 theJob wrote:On September 09 2012 03:15 Anomi wrote:On September 09 2012 01:56 theJob wrote:
There seem to be alot of people here proclaiming that the trend of foreigners’ winratio is negatively related to time. After a quick glance I’m inclined to agree, however if we are to indulge ourselves with using a technical term like “trend” then I feel like a more thorough analysis is due.
If one were to fit an OLS regression on the material it is obvious that the trend is negative. A linear regression tells us that the relationship between the win percentage of foreigners versus Korean players is declining with roughly 1,5 % per quarter. The regression holds true on a 95% significance level and has an explanationvalue of 0,55 R^2.
However after investigating the histogram further it should become obvious that if we want to fit the most accurately representative trend line to the data we have to step away from a linear relation between the two variables. When running a quadratic OLS regression we get much better results than the linear one. There is a statistically significant relation on a 99% significance level and the R^2 value is as high as 0,90. By looking at the scatterplot where the quadratic trendline is fitted I’d have to echo esfiworlds statement that foreigners winpercentage was declining rapidly in the beginning but that it has settled or staring to rise even.
So if we are to use a technical term like “trend”, then the data from esfiworld shows that foreigners performance as of late is improving. I am not wrong s data cant be used for anything statistical at all. The op decided to exclude observations. Even if he thinks that there are less top players playing below top 16 its still an opinion and should not effect the sample size. We all know for instance that stephano have played polt in the pool play. IPL hade a pol play that was really Korean heavy . Allot of observation have been excluded so doing a osl of any sort is out of the question (not saying your analyse is wrong but the data u are using is not good). There allot more issues with the data itself that i don't feel like rambling up. To keep it short each tournaments win ratio is repented by diffrent variabels effecting the outcome. They are not the same so i belive u cant add them up. My post touched on the notion that the material shows a negative trend (which it doesn't necessarily do). I never intended to prove or disprove that the sample data is appropriate for the OPs hypothesis. So yes you are right that if one were to not accept the OPs sample for whatever reason, then it’s not interesting what kind of trend there is. But again, I never touched upon that so I'm confused why you bring that up. Your second point seems to be a bit off. As far as I understand it the OP doesn’t intend to present a comprehensive model which explains why players loose or win tournaments but rather to show whether or nor the skill gap between foreigners and Korean players are changing. For this you would simply need a time variable to represent the flow of time and a second variable representing the skill gap. Foreigner vs Korean win percentages seem to be a reasonable demonstration of differences in skill, at least to me. I think u have misinterpreted me. As u can see i only quoted your post since i am well aware that he sated in his post. If it was about the op i would have quoted him . What my post was directed to was your statement alone wich is allot different from the op. " A linear regression tells us that the relationship between the win percentage of foreigners versus Korean players is declining with roughly 1,5 % per quarter. The regression holds true on a 95% significance level and has an explanationvalue of 0,55 R^2. " This interpetation of the result is correct. If we are to use a term like "trend" then the relationship (or the coefficient) between the time variable and the win percentage variable is -1,577, under a linear OLS regression. This model has a R^2 of 0,55 and holds on a 95% significanse level. But I was claiming that a linear trend isn't optimal and that the best fitted line shows a decrease in skillgap recently.
Then i apologies . Seems that i was the one misunderstanding on what u were claiming.
|
On September 09 2012 04:33 Anomi wrote:Show nested quote +On September 09 2012 04:28 theJob wrote:On September 09 2012 03:51 Anomi wrote:On September 09 2012 03:40 theJob wrote:On September 09 2012 03:15 Anomi wrote:On September 09 2012 01:56 theJob wrote:
There seem to be alot of people here proclaiming that the trend of foreigners’ winratio is negatively related to time. After a quick glance I’m inclined to agree, however if we are to indulge ourselves with using a technical term like “trend” then I feel like a more thorough analysis is due.
If one were to fit an OLS regression on the material it is obvious that the trend is negative. A linear regression tells us that the relationship between the win percentage of foreigners versus Korean players is declining with roughly 1,5 % per quarter. The regression holds true on a 95% significance level and has an explanationvalue of 0,55 R^2.
However after investigating the histogram further it should become obvious that if we want to fit the most accurately representative trend line to the data we have to step away from a linear relation between the two variables. When running a quadratic OLS regression we get much better results than the linear one. There is a statistically significant relation on a 99% significance level and the R^2 value is as high as 0,90. By looking at the scatterplot where the quadratic trendline is fitted I’d have to echo esfiworlds statement that foreigners winpercentage was declining rapidly in the beginning but that it has settled or staring to rise even.
So if we are to use a technical term like “trend”, then the data from esfiworld shows that foreigners performance as of late is improving. I am not wrong s data cant be used for anything statistical at all. The op decided to exclude observations. Even if he thinks that there are less top players playing below top 16 its still an opinion and should not effect the sample size. We all know for instance that stephano have played polt in the pool play. IPL hade a pol play that was really Korean heavy . Allot of observation have been excluded so doing a osl of any sort is out of the question (not saying your analyse is wrong but the data u are using is not good). There allot more issues with the data itself that i don't feel like rambling up. To keep it short each tournaments win ratio is repented by diffrent variabels effecting the outcome. They are not the same so i belive u cant add them up. My post touched on the notion that the material shows a negative trend (which it doesn't necessarily do). I never intended to prove or disprove that the sample data is appropriate for the OPs hypothesis. So yes you are right that if one were to not accept the OPs sample for whatever reason, then it’s not interesting what kind of trend there is. But again, I never touched upon that so I'm confused why you bring that up. Your second point seems to be a bit off. As far as I understand it the OP doesn’t intend to present a comprehensive model which explains why players loose or win tournaments but rather to show whether or nor the skill gap between foreigners and Korean players are changing. For this you would simply need a time variable to represent the flow of time and a second variable representing the skill gap. Foreigner vs Korean win percentages seem to be a reasonable demonstration of differences in skill, at least to me. I think u have misinterpreted me. As u can see i only quoted your post since i am well aware that he sated in his post. If it was about the op i would have quoted him . What my post was directed to was your statement alone wich is allot different from the op. " A linear regression tells us that the relationship between the win percentage of foreigners versus Korean players is declining with roughly 1,5 % per quarter. The regression holds true on a 95% significance level and has an explanationvalue of 0,55 R^2. " This interpetation of the result is correct. If we are to use a term like "trend" then the relationship (or the coefficient) between the time variable and the win percentage variable is -1,577, under a linear OLS regression. This model has a R^2 of 0,55 and holds on a 95% significanse level. But I was claiming that a linear trend isn't optimal and that the best fitted line shows a decrease in skillgap recently. Then i apologies . Seems that i was the one misunderstanding on what u were claiming.
np, im sorry if i was unclear in my original post
|
|
|
|
|
|
EPT
