In any field, data is misrepresented by people who fail to understand nuance. Look at climate change. We've faced some of the most extreme weather across globe than has been seen for decades. But because it was called global warming at one point, dumb fucks will say, "according to our data winter wasn't any hotter than last year" whilst hundreds died to heat stroke in the summer.
But LS is wrong in that "stats are useless". No, stats are useless WITHOUT CONTEXT. But stats are not useless. His example was Kassadin. Ok, what is the winrate delta when the opponent has say >60% of magic damage dealt. How about the winrate delta when Kass is against an AP top and mid vs just AP mid. How about vs immobile? Its not useless. It just needs context and he's out of his mind, either dishonest and in bad faith or just ignorant if he thinks data is useless
The data sets used to measure climate change is also very poor as well though is the issue, the margin for error is so big that any statistician would say that its unreliable. Then you look at the green movement in places like Germany where they had to turn coal-fired plants back on because they tried to rely on solar energy, I don't think this is something that we can grapple with, with our current capabilities at least. We can't even accurately measure the impact of our attempts to combat climate change, sure we can measure the ppm in the air but we can't project our datasets outwards without the margin for error being being bigger than the degree to which the dataset predicts the climate will change which is ridiculous. Imagine the margin for error being bigger than the actual change that is predicted itself and were supposed to reshape our policy and economics around that?
3:36 "when someone uses data as the primary thing of the argument, they are clueless" Absolutely true. Context is so important and that's why data can be misinterpreted
It's sad to see such a necessary person for the development of League's game strategy and knowledge to, effectively, give up. It may have to do with his mental struggles outside of League as well and I can fully empathize with LS feeling like the work and effort he puts in ends up unappreciated and scrutinized. I can see why he would give up, especially after the C9 incident, but I wish that his mental health takes a positive turn sometime in the future and LS will decide to write something like a 100 page LS manifesto about current state of League and his views as his last attempt of trying to make a difference (or something of a similar nature).
My Econ profs would have laughed at these pseudo-analyses. It is ridiculous to just simply look at data without inferring a hypothesis first, an analytical reasoning that precedes analysis and would be proved by data. Phreak's video where he uses 95% CI intervals is a massive nothing-burger - to analyze WR in such a way you would need to assume that all picks are picked blindly with no regard to enemy + own team comps, champs are picked based solely on viability and not personal knowledge/skill of the champ, etc. The average is just an average, easily skewed, and says nothing about the individual viability of the champ without controlling any other variables.
I lose my mind when people do bad science. I've messaged LS about how astronomically bad the stats are in the league community and no one is holding them accountable. A well known example was Phreak's hitjob on LS during the Worlds tierlist debate where Phreak compared the win rates of several jungles and how they would perform in mid (Lillia and Rumble). There's no way you can compare champions without ANOVA and Phreak misrepresented the R-value for variance in the data, in comparison to some trend he has on google docs, to mean that his statistics were proving him right. This hitjob video still exists on Phreak's channel, it's insane how bad the stats are in league and how people are getting "analyst" jobs and credentials without taking a single stats class.
@@fr1iend let's say you were comparing the winrates of champions (with data from different ranks and regions), you would have variables like Ryze, Leblanc or Azir. They would all have means and variance, on a boxplot you could see the percentiles (IQR) and start making assumptions on their differences. However, without checking through ANOVA, we cannot discern if the differences were due to pure luck, resulting in a false positive, or actually different. ANOVA tests, gives us significant power, the ability to test for false positives for the entire dataset and between variables. So if the mean winrate was 50 +- 0.7 and azir was 52 +-0.2, ANOVA might say Azir is significantly different. 50.5 +- 0.1 would not be. 51 +- 0.08? Not likely. 51.2 +- 0.06? Maybe, this requires ANOVA.
@@alanc6468 im wondering how anova could be this powerful and check for false positives and luck? What would determine a lucky win vs champion power based win?
@@lekken560 Power is a term in stats for identifying the likelihood the results are a false positive. It's based on probabilities and distributions, usually we want the percent chance that the results are a false positive to be lower than 5%. We won't know for sure if a win was pure luck or due to draft but we would have a good idea of how likely the results are due to chance. If the p-value was 0.01, we would say that there was a 1% chance of a better result, the result is a part of the upper echelon of results (99th percentile). ANOVA is just a good statistics calculation, you could do it through a calculator (but that require a lot of hand work). We need ANOVA because other tests would not be accurate in determining the chance of luck.
I have been a top few hundreds in an mmo before(with niche world record runs). One of the occurrences that happens once in a while around me is. I will walk into a new raid, take over the lead role, demonstrate why the guides and websites they used are shit, actually clearing it with better efficiency, the previous lead rage quit the guild. Some people attached their identity to the data and guides, the moment i break those down, they break down. Very sad.
Abt what he says at 00:50, it is true for league with it being as broad as it is, in more intimate forms of competition though (chess or card/fighting games) its precisely that kind of information that makes the difference between the L and W
Honestly LS should probably just slowly transition out of league and into other games that he enjoys more. He has the skills to climb back up as a streamer and he probably has more than enough money to not even care that much about success anymore. There is a difference between giving up and letting go.
Eh he says data is useless, and then immediately explains why it would be useful. He really means it cant be taken at face value. Like with the rammus example, what he was essentially getting at is that if a counter-pick type champion is at a 50% wr they are busted in the right scenario but if a first-pick type champion is at a 50% wr they are probably balanced. So rammus being at 50% wr IS useful information if you interpret it properly with context.
LS kinda doesnt understand data analysis which i assume is because he hasnt experienced how people do it properly in research and businesses. Data analysis is always about finding patterns in the data you have and its obviously impossible to find innovation though data(but it may be possible to find pointers on where to look). However stuff like LSTM networks(see dota 2) can come up with new data that can be analysed though ML. Some examples(e.g. supercomputers??) are also just wrong. Idk i dont think hes educated enough to actually speak about this productively outside of the application he has seen in league However the takeaway tldr of "dont use data without context" is basically the fundermental principle of data analysis. EDIT: Im kinda confused by the kassadin and rammus example because those can be spotted though data by looking at winrates vs specific matchup pools.
Yeah, he's complaining about bad data and analysis and the reasons he gives are based on better/newer data but he calls it "information" instead. The 2010 Starcraft player has a better understanding of the game because over time more data has been analyzed and play-styles changed based on that. Picking Rammus into anything is dumb because people are using a very basic dataset to analyze his win rate. Extending that dataset, to factor in matchups and team comps, will give you better data and a better understanding of when rammus is a good pick. I think he's trying to say that intuition is more important, but a lot of the time intuition is based on past experience. A veteran player who's played more games will better understand good/bad match ups because he's experienced them, but just because that information is in his head, instead of on a spreadsheet, doesn't mean it isn't data.
@@pearce05 You guys are still both missing the point. Obviously he understands the argument about scope, he's not saying that is the sole reason stats have limited utility compared to more hardcore analysis--LS is just giving a simple example of how stats are often misinterpreted. The point about incomplete information is not in reference to a missing statistical data point. It's that the statistics themselves are inherently limited by the context they are in. Example: Tic tac toe meta--The original tic tac toe meta states that X is completely overpowered because it goes first. Everyone including professional Tic tac toers agree! In fact O has 0% winrate and loses every game. What we have to ask ourselves, when determining whether or not X or O has a "real" advantage, has nothing to do with the statistics... yet this is what the League community reaches for, first and foremost. League players say, "X has 100% winrate--it's broken!" Statistics do nothing but synthesize what we already know into a mathematical format. They say basically nothing regarding whether what we know about that subject is right or wrong at a deep level. For this Tic tac toe example, the way we draw a bonafide conclusion, involves rather understanding the rules of the game, analyzing the decisions of both X and O, and finally analyzing the possible, yet unmade decisions, of X and O. These possible, yet unmade decisions, are precisely what LS is referring to when he talks about "information." In this case, upon analysis, players with X and O simply raced independently in their own respective vertical lines, leading to a situation where the player going first would always win. Understanding the rules of tic tac toe, we can safely say that O is not actually underpowered--their strategy sucks, they don't understand the game, nor do they understand the real reason they have 0% winrate. X also has no reason to adapt their strategy, it works every time. Actual progression in the tic tac toe space would come from this sort of analysis, not saying pick X every time you can, and play like this, even if it works. Ultimately, LS in my opinion, is someone who understands that we don't know everything, values questioning our underlying assumptions (which are baked in to any statistic), and pushing for real progression in the League space, by analyzing and expanding on what we know or believe to be true.
In any field, data is misrepresented by people who fail to understand nuance. Look at climate change. We've faced some of the most extreme weather across globe than has been seen for decades. But because it was called global warming at one point, dumb fucks will say, "according to our data winter wasn't any hotter than last year" whilst hundreds died to heat stroke in the summer.
It's easy to present data, but to explain it is in an another realm of complexity
its because the general public is stupid but yet they somehow have power
But LS is wrong in that "stats are useless". No, stats are useless WITHOUT CONTEXT. But stats are not useless. His example was Kassadin. Ok, what is the winrate delta when the opponent has say >60% of magic damage dealt. How about the winrate delta when Kass is against an AP top and mid vs just AP mid. How about vs immobile? Its not useless. It just needs context and he's out of his mind, either dishonest and in bad faith or just ignorant if he thinks data is useless
@@crymoreyes if you're typing this then you clearly did not watch or listen to anything he said
The data sets used to measure climate change is also very poor as well though is the issue, the margin for error is so big that any statistician would say that its unreliable. Then you look at the green movement in places like Germany where they had to turn coal-fired plants back on because they tried to rely on solar energy, I don't think this is something that we can grapple with, with our current capabilities at least. We can't even accurately measure the impact of our attempts to combat climate change, sure we can measure the ppm in the air but we can't project our datasets outwards without the margin for error being being bigger than the degree to which the dataset predicts the climate will change which is ridiculous. Imagine the margin for error being bigger than the actual change that is predicted itself and were supposed to reshape our policy and economics around that?
3:36 "when someone uses data as the primary thing of the argument, they are clueless"
Absolutely true. Context is so important and that's why data can be misinterpreted
It's sad to see such a necessary person for the development of League's game strategy and knowledge to, effectively, give up. It may have to do with his mental struggles outside of League as well and I can fully empathize with LS feeling like the work and effort he puts in ends up unappreciated and scrutinized. I can see why he would give up, especially after the C9 incident, but I wish that his mental health takes a positive turn sometime in the future and LS will decide to write something like a 100 page LS manifesto about current state of League and his views as his last attempt of trying to make a difference (or something of a similar nature).
My Econ profs would have laughed at these pseudo-analyses. It is ridiculous to just simply look at data without inferring a hypothesis first, an analytical reasoning that precedes analysis and would be proved by data. Phreak's video where he uses 95% CI intervals is a massive nothing-burger - to analyze WR in such a way you would need to assume that all picks are picked blindly with no regard to enemy + own team comps, champs are picked based solely on viability and not personal knowledge/skill of the champ, etc. The average is just an average, easily skewed, and says nothing about the individual viability of the champ without controlling any other variables.
I lose my mind when people do bad science. I've messaged LS about how astronomically bad the stats are in the league community and no one is holding them accountable. A well known example was Phreak's hitjob on LS during the Worlds tierlist debate where Phreak compared the win rates of several jungles and how they would perform in mid (Lillia and Rumble). There's no way you can compare champions without ANOVA and Phreak misrepresented the R-value for variance in the data, in comparison to some trend he has on google docs, to mean that his statistics were proving him right. This hitjob video still exists on Phreak's channel, it's insane how bad the stats are in league and how people are getting "analyst" jobs and credentials without taking a single stats class.
Sorry data noob here. What is ANOVA?
@@fr1iend let's say you were comparing the winrates of champions (with data from different ranks and regions), you would have variables like Ryze, Leblanc or Azir. They would all have means and variance, on a boxplot you could see the percentiles (IQR) and start making assumptions on their differences. However, without checking through ANOVA, we cannot discern if the differences were due to pure luck, resulting in a false positive, or actually different. ANOVA tests, gives us significant power, the ability to test for false positives for the entire dataset and between variables. So if the mean winrate was 50 +- 0.7 and azir was 52 +-0.2, ANOVA might say Azir is significantly different. 50.5 +- 0.1 would not be. 51 +- 0.08? Not likely. 51.2 +- 0.06? Maybe, this requires ANOVA.
@@alanc6468 im wondering how anova could be this powerful and check for false positives and luck? What would determine a lucky win vs champion power based win?
@@lekken560 Power is a term in stats for identifying the likelihood the results are a false positive. It's based on probabilities and distributions, usually we want the percent chance that the results are a false positive to be lower than 5%. We won't know for sure if a win was pure luck or due to draft but we would have a good idea of how likely the results are due to chance. If the p-value was 0.01, we would say that there was a 1% chance of a better result, the result is a part of the upper echelon of results (99th percentile). ANOVA is just a good statistics calculation, you could do it through a calculator (but that require a lot of hand work). We need ANOVA because other tests would not be accurate in determining the chance of luck.
I have been a top few hundreds in an mmo before(with niche world record runs). One of the occurrences that happens once in a while around me is. I will walk into a new raid, take over the lead role, demonstrate why the guides and websites they used are shit, actually clearing it with better efficiency, the previous lead rage quit the guild. Some people attached their identity to the data and guides, the moment i break those down, they break down. Very sad.
Abt what he says at 00:50, it is true for league with it being as broad as it is, in more intimate forms of competition though (chess or card/fighting games) its precisely that kind of information that makes the difference between the L and W
LS aka 'loathes statistics' does not care that 99% of people pronounce champion names differently than he does
Honestly LS should probably just slowly transition out of league and into other games that he enjoys more. He has the skills to climb back up as a streamer and he probably has more than enough money to not even care that much about success anymore.
There is a difference between giving up and letting go.
Does anybody know hoiw much permanent visa in south korea costs?
just interested how much can something like that be...
Eh he says data is useless, and then immediately explains why it would be useful. He really means it cant be taken at face value. Like with the rammus example, what he was essentially getting at is that if a counter-pick type champion is at a 50% wr they are busted in the right scenario but if a first-pick type champion is at a 50% wr they are probably balanced. So rammus being at 50% wr IS useful information if you interpret it properly with context.
When was this stream?
August 14. Link in description.
bro were are the liveview for this year
No more videos I'm p sure :(
That sucks cause he's not uploading all the games@@EravohnRname
Yeah but sadly he doesn't want me to and I'll respect that
LS kinda doesnt understand data analysis which i assume is because he hasnt experienced how people do it properly in research and businesses. Data analysis is always about finding patterns in the data you have and its obviously impossible to find innovation though data(but it may be possible to find pointers on where to look). However stuff like LSTM networks(see dota 2) can come up with new data that can be analysed though ML.
Some examples(e.g. supercomputers??) are also just wrong. Idk i dont think hes educated enough to actually speak about this productively outside of the application he has seen in league
However the takeaway tldr of "dont use data without context" is basically the fundermental principle of data analysis.
EDIT: Im kinda confused by the kassadin and rammus example because those can be spotted though data by looking at winrates vs specific matchup pools.
Yeah, he's complaining about bad data and analysis and the reasons he gives are based on better/newer data but he calls it "information" instead. The 2010 Starcraft player has a better understanding of the game because over time more data has been analyzed and play-styles changed based on that. Picking Rammus into anything is dumb because people are using a very basic dataset to analyze his win rate. Extending that dataset, to factor in matchups and team comps, will give you better data and a better understanding of when rammus is a good pick.
I think he's trying to say that intuition is more important, but a lot of the time intuition is based on past experience. A veteran player who's played more games will better understand good/bad match ups because he's experienced them, but just because that information is in his head, instead of on a spreadsheet, doesn't mean it isn't data.
@@pearce05 You guys are still both missing the point. Obviously he understands the argument about scope, he's not saying that is the sole reason stats have limited utility compared to more hardcore analysis--LS is just giving a simple example of how stats are often misinterpreted. The point about incomplete information is not in reference to a missing statistical data point. It's that the statistics themselves are inherently limited by the context they are in.
Example:
Tic tac toe meta--The original tic tac toe meta states that X is completely overpowered because it goes first. Everyone including professional Tic tac toers agree! In fact O has 0% winrate and loses every game. What we have to ask ourselves, when determining whether or not X or O has a "real" advantage, has nothing to do with the statistics... yet this is what the League community reaches for, first and foremost. League players say, "X has 100% winrate--it's broken!" Statistics do nothing but synthesize what we already know into a mathematical format. They say basically nothing regarding whether what we know about that subject is right or wrong at a deep level. For this Tic tac toe example, the way we draw a bonafide conclusion, involves rather understanding the rules of the game, analyzing the decisions of both X and O, and finally analyzing the possible, yet unmade decisions, of X and O. These possible, yet unmade decisions, are precisely what LS is referring to when he talks about "information." In this case, upon analysis, players with X and O simply raced independently in their own respective vertical lines, leading to a situation where the player going first would always win. Understanding the rules of tic tac toe, we can safely say that O is not actually underpowered--their strategy sucks, they don't understand the game, nor do they understand the real reason they have 0% winrate. X also has no reason to adapt their strategy, it works every time. Actual progression in the tic tac toe space would come from this sort of analysis, not saying pick X every time you can, and play like this, even if it works. Ultimately, LS in my opinion, is someone who understands that we don't know everything, values questioning our underlying assumptions (which are baked in to any statistic), and pushing for real progression in the League space, by analyzing and expanding on what we know or believe to be true.