|
Personal attacks in this thread will draw a temp ban. |
On March 16 2012 22:15 vlkv wrote:+ Show Spoiler +
The use of statistics in Starcraft 2 keeps annoying me. As Greg once said, quoting stats like that is for people who don't understand the game. And while I don't claim to understand the game, I understand enough about statistics to recognize the misuse of them.
This are some of the problems:
1. Quoting win percentages for 2012 is arbitrary. Why 2012? Why not since December 2011? Or 2011 as a whole? Sure, it is supposed to demonstrate the most recent trends. The question arises though, whether people then make predictions about the very near future (where such numbers might have some meaning), or for a longer period of time. The latter is not a viable use of such statistics, however this is practiced regularly. People in this very thread even went as far as to make predictions about Idra's whole career based on the few performances shown in the last couple of weeks. Needless to say this is completely dumb.
2) When people come in here and make broad statements like "Idra is not Code S material" then I don't have a problem with that unless they decide to back this statement up with ridiculous statistics like under 1). The reason is that this quote implies stability. It is like saying that Idra's talents are not suited for Code S or that he simply does not have enough of trait X to succeed in such competition. To back up such an implicit claim with empirical data would necessitate the use of information that goes back to the beginning of Starcraft 2 and maybe even Idra's career in Broodwar, at least if one accepts that the basic, moderately to highly stable traits necessary to succeed in both games (perception skills, motor skills, intelligence, motives, self-control etc.) are very similar. Needless to say, Idra has had a number of meaningful achievements as a progamer and past success predicts future success. Obviously, environments change, the game changes and all of these things influence the performance of player X with his idiosyncratic skill set. However, people never think about whether these fluctuations in conditions have meaningful stability over time. Which brings me to 3).
3) "Numbers do not lie", they say. Well, apart from the fact that many stats are made-up, are not representative or keep being misused, numbers DO not lie. However, they don't say something in the first place. Statistics are subject to interpretation. And without someone capable and qualified to do this interpretation, they are meaningless. If all you have to bring to this thread or to the community is a win percentage derived from an unrepresentative time sample, without making a point that explains these numbers, and then make broad statements and predictions about long-term effects, you are an idiot and no-one should listen to you. Unfortunately, many people these days are idiots.
4) "Training in Korea did not pay off", they say. This is a horrible brain-fart. In its logical fallacy it is even more severe than 1) and 2). How can one falsify such a statement? Do we somehow know how Idra would perform had he NOT been in Korea for months to practice? Maybe it would have been worse than it is now - if that was true, the quoted statement would be wrong. However, we do NOT know how he would perform. This statement is therefore meaningless. It is merely an opinion, it is not even something that can be debated. Needless to say it also assumes that periods of intense training have an immediate effect. Many progamers claim the opposite. But also athletes, at the advanced and elite levels, are aware of delayed training effects. They often explain, that the improvement in skill is masked by fatigue in periods of very intense and frequent physical stress. To show their true skill, they have a period pre competition, where frequency or other parameters of training are heavily reduced so that fatigue gets eliminated and not any longer suppresses the true, enhanced skill. I am not saying that this is the case for Idra or other gamers. I am merely providing another interpretation of the lack of success in Idra's competition post-Korea, that brings a different perspective into this discussion. Also, it hopefully lets people making such statements look more stupid. I would enjoy that greatly.
I read the entire post, and I agree that it's worth thinking about, but I do not agree with the entirety of it.
Statistics is just another form of evidence. I do not believe that statistics alone suffices to be an argument, and I definitely do not agree that it is always correct, yet they do form a part of an argument, and I'd be sorry if arguments omitted good uses of statistics because that robs us of a source of evidence.
As regards the statistics quotes in this thread (and other notable threads on TL at the moment), I agree that there is abuse and misuse, but the only way one is going to get at the bottom of things is through looking into the arguments that use statistics, and trying to find errors in thinking. As you pointed out yourself, it's generally not difficult to find problems. But not all problems damn the entire use of the statistics, and some information can be extracted. This information is valuable because it's a new source to learn from. So, instead of attacking the use of statistics as such, we should all pay more attention to statistics, and see if the better argument wins.
I'll add a new argument against statistics, though. For sc2, they are almost always outdated. The problem is not using arbitrary periods, it's generally needing too long a period to generate a large sample. Due to the constantly shifting meta-game, analysis would be required almost at a weekly level to capture shifts in meta-game, yet the highest level has very few games per week. Furthermore, a lot of the actual meta-shifting occurs in training for which no records are available, such that the statistics is even further behind the actual state of things. This explains why Idra was complaining about Protoss (and citing Nestea as in agreement with him) before the violent downswing in TLPD statistics ever occurred. To assume that Idra is vindicated because of those statistics would commit the same fallacy, though, as one should assume that win rate statistics always have a huge delay, and say little about the current state of the game.
Edit: Son of a fast-typing msjakofsky.
|
On March 17 2012 00:04 Sqorpion wrote:Show nested quote +On March 16 2012 23:54 msjakofsky wrote:+ Show Spoiler +On March 16 2012 22:15 vlkv wrote:
The use of statistics in Starcraft 2 keeps annoying me. As Greg once said, quoting stats like that is for people who don't understand the game. And while I don't claim to understand the game, I understand enough about statistics to recognize the misuse of them.
This are some of the problems:
1. Quoting win percentages for 2012 is arbitrary. Why 2012? Why not since December 2011? Or 2011 as a whole? Sure, it is supposed to demonstrate the most recent trends. The question arises though, whether people then make predictions about the very near future (where such numbers might have some meaning), or for a longer period of time. The latter is not a viable use of such statistics, however this is practiced regularly. People in this very thread even went as far as to make predictions about Idra's whole career based on the few performances shown in the last couple of weeks. Needless to say this is completely dumb.
2) When people come in here and make broad statements like "Idra is not Code S material" then I don't have a problem with that unless they decide to back this statement up with ridiculous statistics like under 1). The reason is that this quote implies stability. It is like saying that Idra's talents are not suited for Code S or that he simply does not have enough of trait X to succeed in such competition. To back up such an implicit claim with empirical data would necessitate the use of information that goes back to the beginning of Starcraft 2 and maybe even Idra's career in Broodwar, at least if one accepts that the basic, moderately to highly stable traits necessary to succeed in both games (perception skills, motor skills, intelligence, motives, self-control etc.) are very similar. Needless to say, Idra has had a number of meaningful achievements as a progamer and past success predicts future success. Obviously, environments change, the game changes and all of these things influence the performance of player X with his idiosyncratic skill set. However, people never think about whether these fluctuations in conditions have meaningful stability over time. Which brings me to 3).
3) "Numbers do not lie", they say. Well, apart from the fact that many stats are made-up, are not representative or keep being misused, numbers DO not lie. However, they don't say something in the first place. Statistics are subject to interpretation. And without someone capable and qualified to do this interpretation, they are meaningless. If all you have to bring to this thread or to the community is a win percentage derived from an unrepresentative time sample, without making a point that explains these numbers, and then make broad statements and predictions about long-term effects, you are an idiot and no-one should listen to you. Unfortunately, many people these days are idiots.
4) "Training in Korea did not pay off", they say. This is a horrible brain-fart. In its logical fallacy it is even more severe than 1) and 2). How can one falsify such a statement? Do we somehow know how Idra would perform had he NOT been in Korea for months to practice? Maybe it would have been worse than it is now - if that was true, the quoted statement would be wrong. However, we do NOT know how he would perform. This statement is therefore meaningless. It is merely an opinion, it is not even something that can be debated. Needless to say it also assumes that periods of intense training have an immediate effect. Many progamers claim the opposite. But also athletes, at the advanced and elite levels, are aware of delayed training effects. They often explain, that the improvement in skill is masked by fatigue in periods of very intense and frequent physical stress. To show their true skill, they have a period pre competition, where frequency or other parameters of training are heavily reduced so that fatigue gets eliminated and not any longer suppresses the true, enhanced skill. I am not saying that this is the case for Idra or other gamers. I am merely providing another interpretation of the lack of success in Idra's competition post-Korea, that brings a different perspective into this discussion. Also, it hopefully lets people making such statements look more stupid. I would enjoy that greatly. implying that the recent 3-4 months' performance of a player is an "arbetrary" argument... sorry dude. it's a good argument. idra isn't doing well now. it doesn't mean less or more than that. attacking the viability of statistics to make your favorite player look good... that's what i would call arbitrary. predictions are of course very subjective, but the simplest logic dictates that people don't make bright predictions of idra. it's much less likely than not, although not impossible that idra will suddenly start winning everything soon. it's not completely dumb, it's like predicting that it will be a rainy day when the sky is filled with dark clouds in the morning.
so if a player has a very bad record in the recent times vs worse competition than koreans, stating that idra is not code s material is ridiculous? it's not ridiculous. it is what common sense dictates. had you stated that incontrol or machine is code s material, everyone would laugh at you, yet their performance is just as bad as idra's recently. what's the difference with idra? it's that he is your favorite player and he had good achievements 1 year ago.
it's really hard to inquire a representative sample size in sc2. one reason is that you need a large sample size. an other reason is that the older the samples are, the less representative they are due to metagame shifting and players getting more skilled overall. when someone uses a time period as an argument, they have to consider both of these and make a compromise. how can you say that this compromise is good or bad? you can't do it and be 100% right. common sense it's all about and that's subjective as hell. common sense dictates that a sample size of 10 games is bad, but it also dictates that a sample size of 1000 games is bad because the first 900 is 100% irrelevant for the current time. but how can you decide that a sample size of 50 is wrong? you can't. you have to accept that it shows something, but it's not completely reliable. just like the cloudy sky in the morning. the statistics will always be a little flawed as an argument, but that's what we have, that's objective and it gives a feel about a players performance.
the point i give you is that numbers aren't always right, and that statistics don't mean we can make 100% accurate predictions. that is absolutely true. but you can't use that as an argument to prove that idra's statistics don't mean that he is not kicking ass now... that would be like stating that a creature with 4 legs that is furry, shaking its tail, woofing, faithful, licking your hands, gnawing on a bone etc is not necessarily a dog. stating stuff like that because you'd love if that dog was in fact a pig is what i would consider as an useless train of thought... and the sad thing is that you'll have people nodding for you because that statement about the dog is in the interest of idra, but that doesn't make it true at all... People went as far as stating IdrAs pro-gamer career was just about done based on the recent 3-4 months results, when in he really trained in Korea, instead of participating in mass tournaments in the past months. Sure, if he had been in tournaments all the time in those months, it could be good statistics, but to some extend when he is just practicing, it's like saying Dimaga is a horrible player and his career is just about over because the only thing in a long time he did was second place in IEM Global Challenge (There are probably a lot of better examples). Also, if people are so confident that he is trash, watch his stream, or IdrA vs WhiteRa on youtube. Those are recent results. Some might say ladder means nothing, but if it means nothing maybe he isn't playing serious and then he's still owning this hard? Even if it's ladder, it shows where his skill level is at. However I would certainly agree with you, if it wasn't for the extend people took it to, and of course the fact that I'm a mega fan of IdrA makes me disagree a little extra 
hmm yea i didn't read through everything just the long post of Vllkv. saying that idra's pro career is over is dumb of course.i know what he's capable of; he beat huk 3-1 recently etc, i think he is a player with good potential. but he can't unleash it.. doesn't matter that his mechanics etc are great if he cant deal with cheese, mind games, etc. the result is the same- losing all the time. he is a good player who's gameplay is very limited because he is extremely good at some parts of the game but can't keep up with the competition at other aspects. like an athlete whose muscles are world class but his lung is in a bad condition. he is not trash of course... i try not to listen to the haters, but not to the lovers either. make it as objective as possible 
|
does anyone have the vods of him and machine casting
myegnet will take forever to upload them
|
On March 17 2012 00:14 msjakofsky wrote:Show nested quote +On March 17 2012 00:04 Sqorpion wrote:On March 16 2012 23:54 msjakofsky wrote:+ Show Spoiler +On March 16 2012 22:15 vlkv wrote:
The use of statistics in Starcraft 2 keeps annoying me. As Greg once said, quoting stats like that is for people who don't understand the game. And while I don't claim to understand the game, I understand enough about statistics to recognize the misuse of them.
This are some of the problems:
1. Quoting win percentages for 2012 is arbitrary. Why 2012? Why not since December 2011? Or 2011 as a whole? Sure, it is supposed to demonstrate the most recent trends. The question arises though, whether people then make predictions about the very near future (where such numbers might have some meaning), or for a longer period of time. The latter is not a viable use of such statistics, however this is practiced regularly. People in this very thread even went as far as to make predictions about Idra's whole career based on the few performances shown in the last couple of weeks. Needless to say this is completely dumb.
2) When people come in here and make broad statements like "Idra is not Code S material" then I don't have a problem with that unless they decide to back this statement up with ridiculous statistics like under 1). The reason is that this quote implies stability. It is like saying that Idra's talents are not suited for Code S or that he simply does not have enough of trait X to succeed in such competition. To back up such an implicit claim with empirical data would necessitate the use of information that goes back to the beginning of Starcraft 2 and maybe even Idra's career in Broodwar, at least if one accepts that the basic, moderately to highly stable traits necessary to succeed in both games (perception skills, motor skills, intelligence, motives, self-control etc.) are very similar. Needless to say, Idra has had a number of meaningful achievements as a progamer and past success predicts future success. Obviously, environments change, the game changes and all of these things influence the performance of player X with his idiosyncratic skill set. However, people never think about whether these fluctuations in conditions have meaningful stability over time. Which brings me to 3).
3) "Numbers do not lie", they say. Well, apart from the fact that many stats are made-up, are not representative or keep being misused, numbers DO not lie. However, they don't say something in the first place. Statistics are subject to interpretation. And without someone capable and qualified to do this interpretation, they are meaningless. If all you have to bring to this thread or to the community is a win percentage derived from an unrepresentative time sample, without making a point that explains these numbers, and then make broad statements and predictions about long-term effects, you are an idiot and no-one should listen to you. Unfortunately, many people these days are idiots.
4) "Training in Korea did not pay off", they say. This is a horrible brain-fart. In its logical fallacy it is even more severe than 1) and 2). How can one falsify such a statement? Do we somehow know how Idra would perform had he NOT been in Korea for months to practice? Maybe it would have been worse than it is now - if that was true, the quoted statement would be wrong. However, we do NOT know how he would perform. This statement is therefore meaningless. It is merely an opinion, it is not even something that can be debated. Needless to say it also assumes that periods of intense training have an immediate effect. Many progamers claim the opposite. But also athletes, at the advanced and elite levels, are aware of delayed training effects. They often explain, that the improvement in skill is masked by fatigue in periods of very intense and frequent physical stress. To show their true skill, they have a period pre competition, where frequency or other parameters of training are heavily reduced so that fatigue gets eliminated and not any longer suppresses the true, enhanced skill. I am not saying that this is the case for Idra or other gamers. I am merely providing another interpretation of the lack of success in Idra's competition post-Korea, that brings a different perspective into this discussion. Also, it hopefully lets people making such statements look more stupid. I would enjoy that greatly. implying that the recent 3-4 months' performance of a player is an "arbetrary" argument... sorry dude. it's a good argument. idra isn't doing well now. it doesn't mean less or more than that. attacking the viability of statistics to make your favorite player look good... that's what i would call arbitrary. predictions are of course very subjective, but the simplest logic dictates that people don't make bright predictions of idra. it's much less likely than not, although not impossible that idra will suddenly start winning everything soon. it's not completely dumb, it's like predicting that it will be a rainy day when the sky is filled with dark clouds in the morning.
so if a player has a very bad record in the recent times vs worse competition than koreans, stating that idra is not code s material is ridiculous? it's not ridiculous. it is what common sense dictates. had you stated that incontrol or machine is code s material, everyone would laugh at you, yet their performance is just as bad as idra's recently. what's the difference with idra? it's that he is your favorite player and he had good achievements 1 year ago.
it's really hard to inquire a representative sample size in sc2. one reason is that you need a large sample size. an other reason is that the older the samples are, the less representative they are due to metagame shifting and players getting more skilled overall. when someone uses a time period as an argument, they have to consider both of these and make a compromise. how can you say that this compromise is good or bad? you can't do it and be 100% right. common sense it's all about and that's subjective as hell. common sense dictates that a sample size of 10 games is bad, but it also dictates that a sample size of 1000 games is bad because the first 900 is 100% irrelevant for the current time. but how can you decide that a sample size of 50 is wrong? you can't. you have to accept that it shows something, but it's not completely reliable. just like the cloudy sky in the morning. the statistics will always be a little flawed as an argument, but that's what we have, that's objective and it gives a feel about a players performance.
the point i give you is that numbers aren't always right, and that statistics don't mean we can make 100% accurate predictions. that is absolutely true. but you can't use that as an argument to prove that idra's statistics don't mean that he is not kicking ass now... that would be like stating that a creature with 4 legs that is furry, shaking its tail, woofing, faithful, licking your hands, gnawing on a bone etc is not necessarily a dog. stating stuff like that because you'd love if that dog was in fact a pig is what i would consider as an useless train of thought... and the sad thing is that you'll have people nodding for you because that statement about the dog is in the interest of idra, but that doesn't make it true at all... People went as far as stating IdrAs pro-gamer career was just about done based on the recent 3-4 months results, when in he really trained in Korea, instead of participating in mass tournaments in the past months. Sure, if he had been in tournaments all the time in those months, it could be good statistics, but to some extend when he is just practicing, it's like saying Dimaga is a horrible player and his career is just about over because the only thing in a long time he did was second place in IEM Global Challenge (There are probably a lot of better examples). Also, if people are so confident that he is trash, watch his stream, or IdrA vs WhiteRa on youtube. Those are recent results. Some might say ladder means nothing, but if it means nothing maybe he isn't playing serious and then he's still owning this hard? Even if it's ladder, it shows where his skill level is at. However I would certainly agree with you, if it wasn't for the extend people took it to, and of course the fact that I'm a mega fan of IdrA makes me disagree a little extra hmm yea i didn't read through everything just the long post of Sqorpion. saying that idra's pro career is over is dumb of course.i know what he's capable of; he beat huk 3-1 recently etc, i think he is a player with good potential. but he can't unleash it.. doesn't matter that his mechanics etc are great if he cant deal with cheese, mind games, etc. the result is the same- losing all the time. he is a good player who's gameplay is very limited because he is extremely good at some parts of the game but can't keep up with the competition at other aspects. like an athlete whose muscles are world class but his lung is in a bad condition. he is not trash of course... i try not to listen to the haters, but not to the lovers either. make it as objective as possible
I agree
|
|
|
thanks watching it right now
so funny lol
|
Does someone know @ what time Machine says something like:
" I casted your game yesterday and it didn't go that well " :D
|
Man he was listening to some really old stuff yesterday. I kinda liked it :D Classic rock gogo
|
|
|
On March 16 2012 23:54 msjakofsky wrote:+ Show Spoiler +implying that the recent 3-4 months' performance of a player is an "arbetrary" argument... sorry dude. it's a good argument. idra isn't doing well now. it doesn't mean less or more than that. attacking the viability of statistics to make your favorite player look good... that's what i would call arbitrary. predictions are of course very subjective, but the simplest logic dictates that people don't make bright predictions of idra. it's much less likely than not, although not impossible that idra will suddenly start winning everything soon. it's not completely dumb, it's like predicting that it will be a rainy day when the sky is filled with dark clouds in the morning.
3 months of accumulated data from competitions might, in general, be a useful proxy for that players' overall skill during that time frame. It is also useful to make predictions about results in the near future. However, if one wants to wish to predict what might happen half a year from now or during the period until the first expansion, it is NOT particularly valid data. To make predictions about long-term effects you would be wise to consider accumulated, relevant information of the whole career span of that player. Why is that?
Athletes as well as gamers experience ups and downs due to a number of influences. Naturally, a person like Idra, who is under close supervision of the community and for which said community used to hold incredibly high standards and expectations, will be shown in a very negative light whenever he is on a losing streak. This is just because this player is so present in the minds of the public, that he cannot lose a game without a large percentage of that public jumping to conclusions about "his mindset", "his practice", "his motivation", etc. This process alone leads to a bias in data selection. Whenever Idra performs well, there are people who will generate statistics that show him winning everything and therefore people will expect him to own GSL immediately. By the same token, once he loses a couple of games people say "Idra's career is over". And why do they do this? Because they do not understand that the data selection process in regards to WHEN data is selected, is biased. Data is not selected by a computer, that decides on random 3 months of performance to analyze. Data is selected by people, fans and critics exactly when a player loses a lot or wins a lot. Naturally, data gathered that way, is full of "error" when the thing one wants to measure is the skill of a player. So, to come back to the first sentence of this answer, "in general" does not apply here.
On March 16 2012 23:54 msjakofsky wrote:+ Show Spoiler +so if a player has a very bad record in the recent times vs worse competition than koreans, stating that idra is not code s material is ridiculous? it's not ridiculous. it is what common sense dictates. had you stated that incontrol or machine is code s material, everyone would laugh at you, yet their performance is just as bad as idra's recently. what's the difference with idra? it's that he is your favorite player and he had good achievements 1 year ago.
Common sense has never been particularly great at understanding statistics and its implications. Moreover, one cannot expect "common sense" to be equipped with the necessary knowledge and experience to judge a players skill and performance in absence of game results. What I mean with that is greatly made visible by commentators' and fans' recent statements about Idra's "poor MLG Winter Arena performance". What most of those people MEAN with that is "poor MLG Winter Arena RESULTS" - because they are no position to actually judge the amount of skill displayed by Idra and how it compares to his skill in the time frame before he went to Korea for example.
About my own bias: It is certainly there since I like Idra a lot as a public persona as well as Starcraft player. However, I have this discussion more because I am disgusted by the stupidity of a large percentage of the community and also some of its public figures, not because Idra would need my defense.
On March 16 2012 23:54 msjakofsky wrote:+ Show Spoiler +it's really hard to inquire a representative sample size in sc2. one reason is that you need a large sample size. an other reason is that the older the samples are, the less representative they are due to metagame shifting and players getting more skilled overall. when someone uses a time period as an argument, they have to consider both of these and make a compromise. how can you say that this compromise is good or bad? you can't do it and be 100% right. common sense it's all about and that's subjective as hell. common sense dictates that a sample size of 10 games is bad, but it also dictates that a sample size of 1000 games is bad because the first 900 is 100% irrelevant for the current time. but how can you decide that a sample size of 50 is wrong? you can't. you have to accept that it shows something, but it's not completely reliable. just like the cloudy sky in the morning. the statistics will always be a little flawed as an argument, but that's what we have, that's objective and it gives a feel about a players performance.
My argument is that game results can be explained by moderately stable player characteristics, environmental circumstances (stable as well as unstable) and the interaction of the two. If one accepts this logic it becomes apparent that consideration of data over a larger period of time would lead to better predictions. That is because using data from a short period of time artificially limits the variance of (unstable) environmental factors.
To illustrate this: If one would only consider a month worth of data, where the meta game is heavily favored towards one race, predicting the long-term future performance of a player from another race, becomes highly unreliable and therefor invalid. That is because although this person's game results were bad during that month, this might change drastically with the natural shifts in meta game over a longer period of time. Realize that meta game is one of very many environmental factors that change over time. I picked it as an example because everyone should be familiar with what it can mean for individual performance.
On March 16 2012 23:54 msjakofsky wrote:+ Show Spoiler +the point i give you is that numbers aren't always right, and that statistics don't mean we can make 100% accurate predictions. that is absolutely true. but you can't use that as an argument to prove that idra's statistics don't mean that he is not kicking ass now... that would be like stating that a creature with 4 legs that is furry, shaking its tail, woofing, faithful, licking your hands, gnawing on a bone etc is not necessarily a dog. it's true of course technically but not logical. stating stuff like that because you'd love if that dog was in fact a pig is what i would consider as an useless train of thought... and the sad thing is that you'll have people nodding for you because that statement about the dog is in the interest of idra, but that doesn't make it logical or sensible at all...
Those things "you give me" were not even my point. I hope you understand my post better after reading this reply though.
|
On March 17 2012 00:10 Ghanburighan wrote:+ Show Spoiler +I read the entire post, and I agree that it's worth thinking about, but I do not agree with the entirety of it.
Statistics is just another form of evidence. I do not believe that statistics alone suffices to be an argument, and I definitely do not agree that it is always correct, yet they do form a part of an argument, and I'd be sorry if arguments omitted good uses of statistics because that robs us of a source of evidence.
As regards the statistics quotes in this thread (and other notable threads on TL at the moment), I agree that there is abuse and misuse, but the only way one is going to get at the bottom of things is through looking into the arguments that use statistics, and trying to find errors in thinking. As you pointed out yourself, it's generally not difficult to find problems. But not all problems damn the entire use of the statistics, and some information can be extracted. This information is valuable because it's a new source to learn from. So, instead of attacking the use of statistics as such, we should all pay more attention to statistics, and see if the better argument wins.
I'll add a new argument against statistics, though. For sc2, they are almost always outdated. The problem is not using arbitrary periods, it's generally needing too long a period to generate a large sample. Due to the constantly shifting meta-game, analysis would be required almost at a weekly level to capture shifts in meta-game, yet the highest level has very few games per week. Furthermore, a lot of the actual meta-shifting occurs in training for which no records are available, such that the statistics is even further behind the actual state of things. This explains why Idra was complaining about Protoss (and citing Nestea as in agreement with him) before the violent downswing in TLPD statistics ever occurred. To assume that Idra is vindicated because of those statistics would commit the same fallacy, though, as one should assume that win rate statistics always have a huge delay, and say little about the current state of the game.
Edit: Son of a fast-typing msjakofsky.
I agree with everything you said. I think using empirically gathered data is one of the most important ways to help generate knowledge. However, gathering data alone does not suffice and gathering biased data usually does not lead to understanding. Data gathering and statistics can be great tools, but it has to be done correctly. Making the mistakes I pointed out is not "correctly".
|
3 months of accumulated data from competitions might, in general, be a useful proxy for that players' overall skill during that time frame. It is also useful to make predictions about results in the near future. However, if one wants to wish to predict what might happen half a year from now or during the period until the first expansion, it is NOT particularly valid data. To make predictions about long-term effects you would be wise to consider accumulated, relevant information of the whole career span of that player. Why is that?
and who wants here to make long term predictions? those are absolutely useless. you're saying this because of the guy who said idra's career will be over or what? i don't think anyone could be right with a 6 month prediction, only with luck.
Athletes as well as gamers experience ups and downs due to a number of influences. Naturally, a person like Idra, who is under close supervision of the community and for which said community used to hold incredibly high standards and expectations, will be shown in a very negative light whenever he is on a losing streak. This is just because this player is so present in the minds of the public, that he cannot lose a game without a large percentage of that public jumping to conclusions about "his mindset", "his practice", "his motivation", etc. This process alone leads to a bias in data selection. Whenever Idra performs well, there are people who will generate statistics that show him winning everything and therefore people will expect him to own GSL immediately. By the same token, once he loses a couple of games people say "Idra's career is over". And why do they do this? Because they do not understand that the data selection process in regards to WHEN data is selected, is biased. Data is not selected by a computer, that decides on random 3 months of performance to analyze. Data is selected by people, fans and critics exactly when a player loses a lot or wins a lot. Naturally, data gathered that way, is full of "error" when the thing one wants to measure is the skill of a player. So, to come back to the first sentence of this answer, "in general" does not apply here.
bias in the data selection is due to stupidity or trolling. my advice is not to take those guys seriously. complete record of the past 3-4 months is not biased data selection... also you're trying to make it seem like that looking at the records of a player would be a science or something lol. it's not that complicated. idra is a polarizing figure because of his manners and habits, and he is losing 4 times as much as he is winning in televised games now. he is a programer, and match results matter. it's this simple. his results are bad. it's also not random 3 months. it's the exact opposite of random. it's precisely the recent 3 months. data selection is this simple. you're explaining all this as if people were arguing like "idra's record in may, october and march is bad so he is bad", but theyre not. what everyone mentions is the past 3-4 months which is the most relevant time now also bordered by the new year for an obvious time border. everything you say is possible but people are not doing those things.
sorry dude but i have the feeling that you're kind of saying that idra's bad results are only due to people looking at stats wrong which is ridiculous. you would be perfectly happy with people looking at stats this way if idra had a 80% winrate in 2012.
Common sense has never been particularly great at understanding statistics and its implications. Moreover, one cannot expect "common sense" to be equipped with the necessary knowledge and experience to judge a players skill and performance in absence of game results. What I mean with that is greatly made visible by commentators' and fans' recent statements about Idra's "poor MLG Winter Arena performance". What most of those people MEAN with that is "poor MLG Winter Arena RESULTS" - because they are no position to actually judge the amount of skill displayed by Idra and how it compares to his skill in the time frame before he went to Korea for example.
About my own bias: It is certainly there since I like Idra a lot as a public persona as well as Starcraft player. However, I have this discussion more because I am disgusted by the stupidity of a large percentage of the community and also some of its public figures, not because Idra would need my defense.
these statistics are not that hard to understand and imply. basically it's this: there are wins and losses. one counts the wins and losses over a relevant period of time. there you go, how complicated is that? we're not talking about the economics of a country. it's really simple. i studied statistics but looking at the win rates of a sc2 player is much simpler than the exercises at a college. it's so simple that anyone with common sense can do it. results are not necessarily perfectly correlated with your definition of skill, that's true, but if someone has 20% winrate that means he is doing bad everywhere in the world. and statistics are correlated with my definition of skill, not perfectly, but enough to make it show something. a player shouldn't lose to the same build 15 times like idra does to double reactor helions. a player shouldn't leave whenever he feels like he is at a disadvantage. cuz he is not completely skilled then in my understanding.
My argument is that game results can be explained by moderately stable player characteristics, environmental circumstances (stable as well as unstable) and the interaction of the two. If one accepts this logic it becomes apparent that consideration of data over a larger period of time would lead to better predictions. That is because using data from a short period of time artificially limits the variance of (unstable) environmental factors.
the problem with this is that data from a long time ago is absolutely irrelevant. metagame shifts are impossible to predict. patches are impossible to predict. luck is impossible to predict. blind counters are impossible to predict. players tend to rise and fall miraculously. stephano was a scrub at a time. desrow is training hard for a long long time and he is still bad. violet disappeared and then started owning face. MC was unsuccessful for a long time. how on earth would you want to make predictions based on games from 1 year ago when stuff like these happen? who the hell could have predicted 1 year ago that mothership would be mandatory in lategame pvz, that stephano will be the best zerg foreigner, that incontrol would never have any more tournament results a year later? sorry i don't think your logic is useful. we can only consider smaller periods for predictions since the quality of the game is mandatory to be similar. people have to play the same game to make predictions useful. if the metagame shifts, if a patch comes out, the game is suddenly not the same as 2 months ago. the gameplay 1 year ago was completely different than it is now, and it will be completely different 1 year from now, so it's absolutely useless to consider older games.
To illustrate this: If one would only consider a month worth of data, where the meta game is heavily favored towards one race, predicting the long-term future performance of a player from another race, becomes highly unreliable and therefor invalid. That is because although this person's game results were bad during that month, this might change drastically with the natural shifts in meta game over a longer period of time. Realize that meta game is one of very many environmental factors that change over time. I picked it as an example because everyone should be familiar with what it can mean for individual performance.
well the metagame shifts are impossible to predict so you only proved here that it's meaningless to make long term predictions. we can make guesses and sometimes they will be right at one point of time (about the metagame) but that's it.
Those things "you give me" were not even my point. I hope you understand my post better after reading this reply though.
i know they weren't supposed to be your points but those were the valid points never the less.
somehow in the root of all this talking is your hope that idra once again will do well and that is 100% possible. but can't be calculated and at this point of time not much is pointing in that direction.
|
wow why are there so many WALL OF TEXTS in this thread
cba to read all of these can anyone give brief summary?
|
On March 17 2012 02:48 msjakofsky wrote:+ Show Spoiler +and who wants here to make long term predictions? those are absolutely useless. you're saying this because of the guy who said idra's career will be over or what? i don't think anyone could be right with a 6 month prediction, only with luck.
I agree with some of that. Yet, people still make stupid statements like the guys discussing the end of Idra's career because it seemed appropriate for them to do so based on 2012s results. So we do not have to pretend that those opinions do not exist. They exist and they are what I argued against, or rather, the lack of foundation for those predictions.
On March 17 2012 02:48 msjakofsky wrote:+ Show Spoiler +bias in the data selection is due to stupidity or trolling. my advice is not to take those guys seriously. complete record of the past 3-4 months is not biased data selection... also you're trying to make it seem like that looking at the records of a player would be a science or something lol. it's not that complicated. idra is a polarizing figure because of his manners and habits, and he is losing 4 times as much as he is winning in televised games now. he is a programer, and match results matter. it's this simple. his results are bad. it's also not random 3 months. it's the exact opposite of random. it's precisely the recent 3 months. data selection is this simple. you're explaining all this as if people were arguing like "idra's record in may, october and march is bad so he is bad", but theyre not. what everyone mentions is the past 3-4 months which is the most relevant time now also bordered by the new year for an obvious time border. everything you say is possible but people are not doing those things.
Looking at the record of a player is NOT science - and that is what I have been arguing the whole time. It is anecdotal evidence at best and not suited to derive valid predictions from. Whether looking at the most recent results is biased sampling or not is depending on a lot of things you seem to overlook. But you are not interested in actually learning something since you already "studied statistics in college" and therefore know how things work. Unfortunately you are completely insensitive towards the misuse of stats in many recent comments of fans, critics, casters and so on.
On March 17 2012 02:48 msjakofsky wrote:+ Show Spoiler +sorry dude but i have the feeling that you're kind of saying that idra's bad results are only due to people looking at stats wrong which is ridiculous. you would be perfectly happy with people looking at stats this way if idra had a 80% winrate in 2012.
I am not saying that at all and I would gladly discuss reasons for bad results. However, with the "ideas" people bring to this discussion we cannot even get to a point where that is possible without bordering on the ridiculous. You holding it against me that I like Idra as a player and therefore being biased is somewhat of a cheap shot. I made it clear that I like him. When your last straw in this discussion is to refer to that to discredit my argumentation, then so be it.
On March 17 2012 02:48 msjakofsky wrote:+ Show Spoiler +these statistics are not that hard to understand and imply. basically it's this: there are wins and losses. one counts the wins and losses over a relevant period of time. there you go, how complicated is that? we're not talking about the economics of a country. it's really simple. i studied statistics but looking at the win rates of a sc2 player is much simpler than the exercises at a college. it's so simple that anyone with common sense can do it. results are not necessarily perfectly correlated with your definition of skill, that's true, but if someone has 20% winrate that means he is doing bad everywhere in the world. and statistics are correlated with my definition of skill, not perfectly, but enough to make it show something. a player shouldn't lose to the same build 15 times like idra does to double reactor helions. a player shouldn't leave whenever he feels like he is at a disadvantage. cuz he is not completely skilled then in my understanding.
Is that what you learned in college? What the hell did you study?
And just for the record, I am not denying Idra's lack of wins in recent competition.
On March 17 2012 02:48 msjakofsky wrote:+ Show Spoiler +the problem with this is that data from a long time ago is absolutely irrelevant. metagame shifts are impossible to predict. patches are impossible to predict. luck is impossible to predict. blind counters are impossible to predict. players tend to rise and fall miraculously. stephano was a scrub at a time. desrow is training hard for a long long time and he is still bad. violet disappeared and then started owning face. MC was unsuccessful for a long time. how on earth would you want to make predictions based on games from 1 year ago when stuff like these happen? who the hell could have predicted 1 year ago that mothership would be mandatory in lategame pvz, that stephano will be the best zerg foreigner, that incontrol would never have any more tournament results a year later? sorry i don't think your logic is useful. we can only consider smaller periods for predictions since the quality of the game is mandatory to be similar. people have to play the same game to make predictions useful. if the metagame shifts, if a patch comes out, the game is suddenly not the same as 2 months ago. the gameplay 1 year ago was completely different than it is now, and it will be completely different 1 year from now, so it's absolutely useless to consider older games.
Data gathered a long time ago is not irrelevant when one accepts the idea that performance today and in the future, to a substantial degree, is influenced by similar, moderately stable player characteristics as performance in the past. No one claims that perfect predictions are possible from that - I certainly don't. It is however silly to disregard such information as irrelevant. You say it is irrelevant only because you are incapable or unwilling to think about HOW it is relevant. The future is always difficult to predict, regardless of short-term or long-term and there is always prediction error. I find it however fascinating that your solution to the problem is either
a) not making predictions ("...is impossible to predict") OR
b) ignoring information ("...is absolutely irrelevant").
Again, what did you study?
On March 17 2012 02:48 msjakofsky wrote:+ Show Spoiler +well the metagame shifts are impossible to predict so you only proved here that it's meaningless to make long term predictions. we can make guesses and sometimes they will be right at one point of time (about the metagame) but that's it.
You again are not even getting what the problem is or which point I tried to get across.
On March 17 2012 02:48 msjakofsky wrote:+ Show Spoiler +somehow in the root of all this talking is your hope that idra once again will do well and that is 100% possible. but can't be calculated and at this point of time not much is pointing in that direction.
I find it contradictory that you think it to be possible to calculate the lack of success, but not the success of a player. How is that possible? Other then that, I indeed hope that Idra will do well. But I don't see what that has to do with our discussion.
|
Wanted to note that Idra is an excellent caster! He should do more more more casts.
|
|
|
Idra doing well on ladder atm. Apart from Stephano and Nerchio, he has the highest win rate of any zerg on EU or NA.
|
On March 17 2012 04:50 Micket wrote:
Idra doing well on ladder atm. Apart from Stephano and Nerchio, he has the highest win rate of any zerg on EU or NA.
And he quits early a lot too. Not in a rage sort of way, but rather "this is no longer worth my time even though I could come back and win." It could be higher.
|
On March 17 2012 04:53 Mr Showtime wrote:Show nested quote +On March 17 2012 04:50 Micket wrote:
Idra doing well on ladder atm. Apart from Stephano and Nerchio, he has the highest win rate of any zerg on EU or NA. And he quits early a lot too. Not in a rage sort of way, but rather "this is no longer worth my time even though I could come back and win." It could be higher.
He also played really sloppy when he was commentating. Also, Protoss.
He could be 85% easy. This is a good sign considering last time he was on NA he was in the mid to high 70's in terms of winrate.
|
This was missed between annoying walls of text.
|
|
|
|
|
|
EPT
